THIERRY MONTMERLE and JEAN-CHARLES AUGEREAU · (like the Sun today) vs. the number of stars in star-forming regions leads to a probability argument drawn from observations: nearly

3. Solar System Formation and Early Evolution:

the First 100 Million Years

THIERRY MONTMERLE and JEAN-CHARLES AUGEREAULaboratoire d’Astrophysique de Grenoble, Universite Joseph Fourier, Grenoble, France

(E-mails: [email protected]; [email protected])

MARC CHAUSSIDONCentre de Recherches Petrographiques et Geochimiques (CRPG), Nancy, France

(E-mail: [email protected])

MATTHIEU GOUNELLE1,2

1Museum National d’Histoire Naturelle, Paris, France; 2Natural History Museum, London, UK


BERNARD MARTYEcole Nationale Superieure de Geologie, Nancy, France


ALESSANDRO MORBIDELLIObservatoire de la Cote d’Azur, Nice, France


(Received 1 February 2006; Accepted 4 April 2006)

Abstract. The solar system, as we know it today, is about 4.5 billion years old. It is widely believed

that it was essentially completed 100 million years after the formation of the Sun, which itself took less

than 1 million years, although the exact chronology remains highly uncertain. For instance: which, of

the giant planets or the terrestrial planets, formed first, and how? How did they acquire their mass?

What was the early evolution of the ‘‘primitive solar nebula’’ (solar nebula for short)? What is its

relation with the circumstellar disks that are ubiquitous around young low-mass stars today? Is it

possible to define a ‘‘time zero’’ (t0), the epoch of the formation of the solar system? Is the solar system

exceptional or common? This astronomical chapter focuses on the early stages, which determine in large

part the subsequent evolution of the proto-solar system. This evolution is logarithmic, being very fast

initially, then gradually slowing down. The chapter is thus divided in three parts: (1) The first million

years: the stellar era. The dominant phase is the formation of the Sun in a stellar cluster, via accretion of

material from a circumstellar disk, itself fed by a progressively vanishing circumstellar envelope. (2) The

first 10 million years: the disk era. The dominant phase is the evolution and progressive disappearance

of circumstellar disks around evolved young stars; planets will start to form at this stage. Important

constraints on the solar nebula and on planet formation are drawn from the most primitive objects in

the solar system, i.e., meteorites. (3) The first 100 million years: the ‘‘telluric’’ era. This phase is

dominated by terrestrial (rocky) planet formation and differentiation, and the appearance of oceans and

atmospheres.

Keywords: Star formation: stellar clusters, circumstellar disks, circumstellar dust, jets and outflows; solar

nebula: high-energy irradiation, meteorites, short-lived radionuclides, extinct radioactivities, supernovae;

planet formation: planetary embryos, runaway growth, giant planets, migration, asteroid belt, formation

of the Moon; early Earth: atmosphere, core differentiation, magnetic field

Earth, Moon, and Planets (2006) � Springer 2006

DOI 10.1007/s11038-006-9087-5

3.1. The First Million Years: The ‘‘Stellar Era’’1

THIERRY MONTMERLE

3.1.1. THE SUN’S BIRTHPLACE

Stars are not born in isolation, but in clusters. This is what astronomicalobservations of our galaxy (the Milky Way) and other galaxies tell us. Thebirthplace of stars are the so-called ‘‘molecular clouds’’, i.e., vast, cold volumesof gas (mostly molecular hydrogen and helium, and also complex organicmolecules, with so far up to 11 C atoms: see Ehrenfreund and Charnley, 2000for a review). These clouds also contain dust grains (which include heavyelements in the form of silicates, hydrocarbons, and various ices). Themasses of molecular clouds typically range from 106 to 108 Mx: in principle,molecular clouds are sufficiently massive to form millions of stars. However,somewhat paradoxically, molecular clouds do not naturally tend to form stars:gravitation, which would tend to generate the ‘‘free-fall’’ collapse of molecularclouds in less than 1 Myr,2 appears to be balanced by an internal source ofpressurewhichkeeps them in gravitational equilibrium.Thebasic answer lies inthe study, in the radio range, of the velocity distribution in the gas. It can beshown that this distribution corresponds to a state of turbulence, i.e., gaseouseddies that exchange energy from the large scale (the size of the cloud) to thesmall scales (‘‘cloudlets’’ of size ~0.1 pc); smaller scales may be present but arecurrently beyond the spatial capabilities of existing radiotelescopes.However, ona large scale, it can be seen frommid- to far-IRobservations (which have a betterspatial resolution than in the mm range), that molecular clouds are in factfilamentary, but these filaments are constantly moving, as attested by theirvelocity distribution. Figure 3.1 shows a 100-micron image by the IRAS satelliteof the Orion molecular cloud complex, where dense and cold filaments areconspicuous. According to turbulence theories (and dedicated laboratoryexperiments), energy is transferred fromthe large scales to the small scales. So thequestion becomes: what drives the turbulence? Or, in other words, where doesthe supporting energy come from?Current explanations are still debated. Theyfocus either on an external energy source like a neighboring supernova, or onaninternal feedbackmechanism: as we shall see below (Section 3.1.2), young starsdrive powerful outflows of matter (Reipurth and Bally, 2001), in such a way

1 How astronomers determine stellar ages is described in Chapter 2 on ‘‘Chronometers’’(Section 2.1).

2 Here we adopt the usual astronomical convention: 1 ‘‘Myr’’ = 106 years. This is exactlysynonymous to 1 ‘‘Ma’’, as used for instance by geologists elsewhere in the article (where ‘‘a’’

stands for ‘‘annum’’).

THIERRY MONTMERLE ET AL.

that they ‘‘inflate’’ the turbulence cells to keep the cloud from collapsing.Nevertheless, energy is dissipated at the smallest scales, so that some form ofcollapse is inevitable: the idea is that the smallest cloud structures, ‘‘cloudlets’’or ‘‘prestellar cores’’, eventually collapse to form stars (e.g., Bate and Bonnell,2004; Goodwin et al., 2004; Padoan et al., 2004).

Other arguments point to an important role of the interstellarmagnetic field.In principle, a molecular cloud is by definition cold and entirely neutral, thuscannot be influenced by the presence of magnetic fields. But in practice, aminute fraction of the gas (roughly 10)7) is ionized (electrically charged) byambient cosmic rays and also by hard radiation from young stars (UV andX-rays, see Section 3.2.1.3). The charged particles are tied to themagnetic field,and thus, through collisions with them, neutral particles are in turn influencedby it: this is called ambipolar diffusion. In a way, ambipolar diffusion acts as a

Figure 3.1. The Orion complex. Left: image of the Orion nebula M42 in the visible domain (�Anglo-Australian Telescope). Background: far-IR image (100 microns) of the Orion complex,by the IRAS satellite (1986), covering a very wide area (the angular scale is given). Note thewidespread filamentary structure of the ‘‘giant molecular cloud’’. The bright spots are several

star-forming regions belonging to the same complex, the most active one being M42 (box).

SOLAR SYSTEM FORMATION AND EARLY EVOLUTION

dragnet through which neutral particles flow across magnetic field lines. Thiseffect is quantitatively important: measurements of magnetic field intensityinsidemolecular clouds (via theZeeman effect onmolecular lines) show that thegas pressure and themagnetic pressure are just about equal, with a difference ofat most a factor of 2 in either direction, depending on the clouds (e.g., Padoanet al., 2004; Crutcher, 2005). Thismeans that in reality we are probably dealingwith magnetically regulated turbulence: the flow of gas in turbulent cells is notfree, but is slowed down by magnetic fields and preferentially proceeds alongfilaments (e.g., Pety andFalgarone, 2003; Falgarone et al., 2005). In particular,this means that, at the small scales, gravitational collapse proceeds either on ashort, free-fall time scale (typically ~104 years) if the magnetic field is on theweak side (magnetic pressure<gaspressure), or on a long, ambipolar diffusiontime scale (which can reach several 105 years or more) if the magnetic fields issufficiently strong (magnetic pressure>gas pressure).

This picture, at least qualitatively, leads to the idea that molecularclouds are stable, self-supported structures, but on the verge of gravita-tional collapse. Depending on the intensity of the magnetic field, starformation may occur at many places in the cloud, perhaps in sequence,within a relatively short timescale, of the order of a few 104 years locally,a few 105 years to 106 years globally. One such global theoretical mech-anism is called ‘‘competitive accretion’’: stars form out of shocks within apool of colliding gaseous filaments, where they compete to acquire theirmass as they move through the cloud (Bate and Bonnell, 2004, Clark et al.2005; Figure 3.2). Pure gravitational collapse has also been advocated(Krumholz et al., 2005).

Whatever the details of the various star formation mechanisms, the netresult is star formation in clusters. Molecular observations (Motte et al.,1998), in the mass range between ~0.1 and 1 Mx, show that above 0.5 Mx

the core mass distribution and the observed stellar ‘‘initial mass function’’(IMF) are the same, which strongly suggests (but does not prove) that thestellar mass distribution directly derives from the core mass distribution,itself linked with the turbulent structure of molecular clouds. (The IMF is thedistribution of stellar masses at formation: it is observed to be a universallaw, expressed as dN*/dlogM* �M*

)1.5 for M* ‡ 0.5 Mx, where N* is thenumber of stars in the mass range M*, M* + dM*; explaining it is one of thehardest challenges for star formation theories: see Kroupa, 2002 for areview.) Depending on a number of external conditions, such as the totalmolecular cloud mass, the passage of a shock wave of a nearby supernovaexplosion from the most massive stars (see below), etc., the high-mass end ofthe IMF is observed to be cut-off at some value Mmax. Some clusters havemassive to very massive stars (Mmax up to several tens of Mx), others haveonly intermediate-mass stars (Mmax = a fewMx at most), all the way to verysmall masses (‘‘brown dwarfs’’, that are not massive enough to eventually


trigger nuclear reactions, M*<0.08 Mx). Because of the observed univer-sality of the IMF, a large Mmax implies a large number of stars, a small Mmax

implies a small number of stars. For example, star-forming regions like Oriondisplay stars up to 20 Mx or more, and contain altogether several thousandstars, while others like Ophiuchus, Taurus, etc., do not go beyond a few Mx

and harbour only a few tens to a few hundred stars.Once the stars are formed, what remains of the parental cloud, not yet

condensed into stars, is eventually dispersed, and the stars become opticallyvisible. At this point, the stellar cluster becomes free from its parent cloud,and its evolution is regulated by dynamical effects in its own gravitationalpotential, leading after a few tens of Myr to ‘‘open clusters’’, then to a broaddispersal of the stars in the galaxy (at typical velocities on the order of a fewkm/s), much like a beehive, and thus to a loss of ‘‘memory’’ of how andwhere they were formed individually.

The Sun probably has been one of such stars. The statistics of ‘‘field stars’’(like the Sun today) vs. the number of stars in star-forming regions leads to aprobability argument drawn from observations: nearly 90% of solar-like starsmust have been born in clusters, of a few tens to a few thousand stars (Adamsand Myers, 2001).

Figure 3.2. Numerical three-dimensional simulation of star formation in a 10,000 Mx cloud,~600,000 yrs after the initial collapse (P.C. Clark, private communication; simulations done at

the UK Astrophysical Fluids Facility in Exeter). The figure is 5 pc on a side. Note thesimilarity of the cloud structure with that of the Orion complex shown in the previous figure.The simulation eventually leads to the formation of ~500 stars. This is less than observed in

Orion (~2000 stars in M42). Complicated effects such as feedback on turbulence from stellaroutflows and ionizing radiation have not been included. A better agreement is expected in thefuture when these effects are taken into account.


It is therefore impossible at present to know a priori in which type ofstellar cluster the Sun was born: our star is about 50 times older than theoldest open clusters. But the issue is fundamental for planet formation ingeneral and for the origin of the solar system in particular, because of theshort time evolution of the circumstellar disks around young stars (see Sec-tion 3.2.1). To simplify, there are two extreme possibilities for the birthplaceof the Sun:(i) The Sun was born in a ‘‘rich’’, Orion-like environment. (see Hillen-

brand 1997; Figure 3.3) The most massive stars (the Trapezium-likestars) are very hot, and thus emit UV photons able to strongly ionizetheir immediate environment. The disks of the less massive stars thentend also to be ionized and evaporate, as shown in Figure 3.4According to most calculations, the disks, which have masses in therange 10)2 – 10)4 MSun (i.e., 0.1–10 Jupiter masses), will disappear ina few million years only, likely before any terrestrial planet has hadthe time to form. One possibility for disks to survive is not to staytoo long in the vicinity of the hottest stars, so as to escape, viadynamical effects, the original ‘‘beehive’’. We observe that most disksaround young stars are typically 10 times larger that the size of thepresent-day solar sytem; some disks may be cut-off to the size of theKuiper Belt (i.e., the radius of the solar sytem, ~50 AU),3 mainlybecause of evaporation processes (Adams et al. 2004, 2006). On theother hand, the discovery of the existence of Sedna, a relativelymassive, high-eccentricity solar system ‘‘planet’’, implies that theprotosolar disk did not suffer stellar encounters closer than 1,000 AU(Morbidelli and Levison, 2004), so retained its initial large size forsome time, or else managed to be not entirely vaporized. This wouldbe possible in the outskirts of the nebula, or in a less rich cluster likeNGC1333 (Adams et al., 2006). In any case, disk survival in a Orion-like environment is certainly not easy.4

(ii) The Sun was born in a ‘‘poor’’, Ophiuchus-like environment (Fig-ure 3.5). With no massive star around, stars form deeply embeddedinside the molecular cloud, and once formed they stay protected fromexternal disturbances. All disks survive, and can remain very large. Weknow of some examples of stars with large disks in the immediatevicinity of molecular clouds (Figure 3.6). It is not clear how in this case

3 1 UA = 1 Astronomical Unit = Sun-Earth distance = 150 million km.

4 To be complete, one should mention the recent work by Throop and Bally (2005), who arguethat dust grains actually grow into planetesimals under the coagulating effect of UV radiation,hence that planet formation (see below, Section 3.2.1), is favored by evaporation. But even in

this case, the exposure to UV radiation must be fine-tuned for the whole system to survive.


Figure 3.3. Near-IR (2 microns) image of the center of M42, revealing the stars of the richOrion nebula cluster, and in particular the four central hot stars called the ‘‘Trapezium’’,which excite the nebula. The nebula is 450 pc away. The image is about 10¢ on a side, which is

250 000 AU. (� ESO, VLT-ISAAC, by M. McCaughrean.).

Figure 3.4. A ‘‘tear’’ in Orion. This is an evaporating circumstellar disk, 500 AU in diameter.The central star is clearly visible. The bright spot is oriented towards h1 Ori C, the hottest starof the Trapezium. The evaporating gas is shaped by the wind from this star (� NASA Hubble

Space Telescope: J. Bally, H. Throop, & C.R. O’Dell).


planet-forming disks would shrink by a factor of 10 to be confinedwithin the size of the solar system. In the end, perhaps the issue of theexistence of a planetary system such as ours is that of the ‘‘survival ofthe fittest’’: to suffer disk truncation in a dense, Orion-like cluster, or atleast to avoid complete evaporation by way of dynamical effects, i.e. tobe ejected from the original cluster quickly enough. The ‘‘quiet’’environment of the Ophiuchus-like, looser clusters, would provide a‘‘safe’’ evolution, but, likely, at the cost of carrying along a large,massive disk which would lead to planetary systems very different fromour own. In this scenario, the solar system would be born in rare,though not uncommon (10% of the stars), environment, with a keyrole played by infrequent dynamical interactions, cutting off its ori-ginal disk, very early (105 years?) after its birth.

3.1.2. THE SUN AS A FORMING STAR

Let us now zoom on the Sun as a forming star, which we shall assumeisolated for simplicity (knowing from above that it has to be isolated, at somevery early stage, from its ‘‘cousins’’ in a cluster of forming stars).

The knowledge of what we believe must have been the first stages offormation and evolution of the Sun, is drawn from a wealth of observationsof a multitude of star-forming regions that astronomers have been able to

Figure 3.5. Left: The ‘‘poor’’ q Oph cluster in the optical range, hidden in a dark cloud of gasand dust. It is located 150 pc from the Sun. (The bright stars at the bottom left is the

foreground star Antares, and the fuzzy spot to its right is the distant globular cluster M15. (�Anglo-Australian Telescope.) Right: Mid-IR image of the cloud core, revealing embeddedyoung stars and protostars invisible in the optical range (ISOCAM, Abergel et al. 1996).


obtain. Nowadays, telescopes are being used both on the ground and inspace, covering almost all the electromagnetic spectrum, from mmwavelengths, to X-rays, across the IR and optical domains. Depending on thewavelength, it is possible to pierce the darkness of molecular clouds, and‘‘see’’ inside them to watch the hidden birth of solar-like stars, mostimportantly in the IR to mm domains, and in the X-ray and gamma-rayranges (e.g., Ryter, 1996).

In almost every case, one is able to distinguish three main components,which simultaneously evolve as star formation proceeds5 (see, e.g., Shu et al.,1987; Andre and Montmerle, 1994; Andre et al., 2000, a summary and recentreferences are given in Feigelson and Montmerle, 1999, and Montmerle,2005: Figure 3.7). At the so-called ‘‘protostellar stage’’, a vast, dense enve-lope (1,000–10,000 AU in radius) is detectable, and from the center emerges a‘‘bipolar outflow’’. The envelope is so dense that its interior is invisible evenat mm wavelengths; only its outer structure can be seen. It is now understoodthat the ‘‘seed’’ of a new star is formed from matter accreted from theenvelope which ‘‘rains’’ on it under the pull of gravitation. The youngestobserved protostars have an estimated age of ~104 years: this estimate is

5 For a pioneering work, establishing (analytically!) the basic principles of early stellar evo-

lution, see Hayashi (1966).

Figure 3.6. A lonely, massive egde-on disk in the ourskirts of the q Oph cloud (circle). The

other disk-like object (dotted circle) is a distant galaxy (� ESO, VLT-ISAAC, Grosso et al.,2003).


rather uncertain, but is consistent with the number deduced from thedynamical age of outflows (= size/velocity), and from their small numberrelative to their more evolved counterparts, like T Tauri stars (see below):indeed, one finds roughly 1 protostar for every 100 T Tauri stars, aged1–10 Myr.

At an age ~105 years, the envelope is much less dense, since most of ithas collapsed onto the disk. It becomes transparent at mm wavelengths,revealing a dense disk (500–1,000 AU in radius), from which the seed starcontinues to grow. The source of outflows become visible, in the form ofhighly collimated jets originating close to the central star, confirmingearlier models in which molecular ourflows consist of cold cloud materialentrained by the jet. (In fact, this jet-cloud interaction is believed by someauthors to be the main agent to sustain the turbulent state of the cloud:this is the ‘‘feedback’’ mechanism mentioned above, see Matzner andMcKee, 2000.) At this stage is revealed the key three-component structurethat governs the physics of star formation: an outer envelope, an inner‘‘accretion’’ disk, and matter ejected perpendicular to the disk (Figure 3.8).(The accretion disk probably exists from the start of collapse, but itcannot be detected because of the opacity of the envelope at the earlieststage.)

Figure 3.7. A summary table of the various protostellar and stellar phases, with characteristic

timescales and basic observational properties. (From Feigelson and Montmerle, 1999).


As time passes, this three-component structure considerably evolves. Theenvelope is eventually exhausted after ~106 years, leaving a fully developed‘‘real’’, luminous star, a massive circumstellar disk (Mdisk ~1–10 MJup), and aweak, optically visible bipolar jet. This is the start of the so-called ‘‘classical’’T Tauri stage (after the name of the first-discovered star of this type, which,as it turns out, is quite atypical for its class; e.g., Bertout, 1989). Then, after aperiod which can be as long as 107 years, the disks of T Tauri stars (and jets)‘‘disappear’’, or at least becomes undetectable, presumably via planet for-mation (see Section 3.2.1): this is the ‘‘weak’’ T Tauri stage. Why ‘‘weak’’?Because the ‘‘classical’’ T Tauri stars have a very unusual optical spectrum,with very strong emission lines, in particular the Ha line of ionized hydrogen;in contrast, the ‘‘weak’’ T Tauri stars have ‘‘weak’’ emission lines, compa-rable to solar lines. A ‘‘weak’’ Ha line can be explained, in analogy with theSun, by the presence of ‘‘active regions’’, that is, starspots scattered over thestellar surface, indicative, again as on the Sun, of locally strong magneticfields. The strong Ha line of ‘‘classical’’ T Tauri stars cannot be explained bymagnetic activity alone. Various arguments assign the strength of the Ha linein this case to matter falling onto the star, fed by the disk: this is the phe-nomenon known as ‘‘accretion’’ of matter (e.g., Bertout, 1989) – the veryprocess by which the star grows to reach its final mass when the disk isexhausted, perhaps leaving young planetary bodies behind.

Let us now concentrate on young ‘‘classical’’ T Tauri stars, like HH30 inTaurus (Figure 3.9). All the observational evidence points to a causal rela-tionship between the existence of jets and the presence of disks: it is clear, atleast qualitatively, that the jet material is somehow coming from the disk, atthe same time that accretion feeds the central star. This phenomenon,known as the ‘‘accretion–ejection’’ phenomenon, is absolutely central to our

Figure 3.8. Sketch of the structure of a protostar, zooming on the star-dik interaction region,

which is dominated by magnetic fields. This region is the seat of the ‘‘accretion–ejection’’mechanism, by which the majority of the disk mass becomes accreted to form a star at thecenter, while the remainder is ejected. (From Feigelson and Montmerle, 1999).


understanding of star formation and early evolution, and to a modern viewof the solar nebula. Such a phenomenon is somewhat paradoxical: it impliesthat, for a star to form, it must lose mass! At least, a significant fraction ofthe mass (>10% from observations, e.g., Muzerolle et al., 2001) musteventually be ejected. The point is that mass accumulation is not the onlynecessity to form a star: since (at we see at all scales in the universe, fromclusters of galaxies to planets) rotation is always present, and seen in the formof circumstellar disks around forming stars, there must exist centrifugalforces that oppose gravity. For gravity to ‘‘win’’ and lead to star formation,angular momentum must be lost. Although the reasons for angular momen-tum loss in a forming star, and in particular the role of the circumstellar disk,are not entirely clear because of complex transport processes within them (seenext subsection), it is well known that mass loss in the form of stellar winds(the solar wind is one example) is very efficient to spin down a star – providedthe ejected matter remains coupled to the star. The only way to do it is to linkthe star and the wind by a magnetic field. Along this line of thought, incurrent models, accretion (mass gain), and ejection (mass and angularmomentum loss), must be mediated by magnetic fields.

3.1.3. A STELLAR VIEW OF THE ‘‘PRIMITIVE SOLAR NEBULA’’

To bemore specific, let us now zoom again, this time on the region very close tothe forming star, for instance a young T Tauri star, to which the primitive solar

Figure 3.9. Left: HST image of the T Tauri star HH30, showing its edge-on disk and jet. Theyoung Sun could have been such an object. Right: sketch of the theoretical magnetic structureused to model the accretion–ejection mechanism (the background drawing is taken from

Ferreira et al., 2000). The ‘‘star’’ is an image of the magnetically active Sun, seen in X-rays bythe Yohkoh satellite (see Section 3.2.1.2).


nebula must have been comparable in its first million years (Figure 3.9, left). Arich interplay between observations and theory, over the last decade, results inthe followingpicture (e.g.,Matsumoto et al., 2000; Shang et al., 2002;Ferreiraand Casse, 2004, and refs therein; see Figure 3.9, right). Both the star and thedisk are magnetized: (i) the star is surrounded by a ‘‘dipolar’’ magnetospherethat surrounds it like a tire, with a ‘‘closed’’, loop-like topology of themagneticfield; (ii) in contrast, the magnetic field lines connected to the disk are ‘‘open’’,above and below the disk. Then a special distance, called the ‘‘corotation ra-dius’’Rc, is naturally defined: this is the distance at which the ‘‘Keplerian’’ (i.e.,orbital) velocity of a disk particle rotates at exactly the same speed as the star.(One could say, in analogy with the Earth’s artificial satellites, that this is the‘‘astrostationary’’ orbit.) At distances r<Rc from the star, the intensity of themagnetic field is stronger than atRc, and themagnetosphere rotates in a ‘‘rigid’’fashion, the field lines being anchored on the stellar surface. At distancesr>Rc, the magnetic field decreases rapidly and takes an open, spiral form as itbecomes tied to the disk. The point at Rc thus has a very particular magneticproperty: it is the border at which the magnetic field topology switches fromclosed (stellar component) to open (disk component). As such, it is also knownas the ‘‘X-point’’ (or more exactly the ‘‘X-ring’’ in three dimensions in view ofthe assumed axial symmetry) because of its X-shaped magnetic configuration(Shu et al. 1997).

The existence of the X-point (in a 2-dimensional cut) holds the key to themajority of ‘‘accretion–ejection’’ theories. There are many discussions amongtheorists about its exact status. For instance, it is not clear why the mag-netically defined X-point (at a distance Rx which depends only on the mag-netic field intensity) should be exactly at the same location as thegravitationally defined corotation radius Rc (which depends only on thestellar mass and rotation velocity), in other words why should Rx = Rc. It isnot clear either that the X-point should be that: a point (or more precisely an‘‘X-ring’’, in three dimensions), since this would mean that at Rx there mustbe an infinite concentration of magnetic fields lines, which is physicallyimpossible if matter is coupled to it (via some ionization, for instance due toX-rays), etc. But most theorists (and observers) agree, at least qualitatively,on the following general ‘‘accretion–ejection’’ picture, which will be sufficientfor the purpose of this paper.

Because of its special gravitational and magnetic properties, seen in twodimensions, the X-point de facto behaves like a Lagrange point (Figure 3.9,right): (i) If a particle, initially located at this point, is pushed towards theinterior (r<Rx), it will start falling freely on the star under the pull ofgravity, along the corresponding ‘‘rigid’’ magnetic field line: this is called‘‘magnetospheric accretion’’. (ii) Conversely, if a particle is pushed outwards,it will start following an open field line, and the centrifugal force will push iteven further: this is how ‘‘centrifugal jets’’ are formed. Thus, the X-point is


intrinsically unstable, analogous to the gravitational Lagrange point L1 be-tween two celestial bodies (which here would be a three-dimensional mag-netic ‘‘Lagrange ring’’ in a star-disk sytem), and flowing through it mattercan either fall onto the star or be ejected. In practice, of course, matter doesboth: depending on the exact model, calculations predict that 1/10 to 1/3 ofthe disk matter should be ejected, the remainder falling on the star and thusfeeding its growth (e.g., Johns-Krull and Gafford, 2002). What is importantfor our purpose is the location of the X-point: for a solar-mass star rotatingrelatively slowly as T Tauri stars are observed to rotate (period of a fewdays), Rx is on the order of several stellar radii. Since a typical T Tauri starhas a radius R* ~ 3 Rx, this means Rx ~ 0.1 AU. In the well-studied T Tauristar AA Tau, for which the accretion disk is seen edge-on, it has been possibleto reconstruct the magnetic structure of the inner hole, from a study of itsactive regions eclipsed by the disk (Bouvier et al., 1999, 2003). Although inthis case the disk is found to be warped, which implies that the magneticaxis is tilted to the rotation axis, the size of the resulting magnetosphere isin good agreement with the theoretical picture (Figure 3.10), in particular

Figure 3.10. Sketch of the tilted magnetic structure inside the disk of the T Tauri star AA Tau.This sketch results from a study of the magnetic activity of the star, eclipsed by its warpedaccretion disk seen nearly edge-on (From Bouvier et al., 1999).


with the presence of an inner hole ~0.1 AU in radius. As a matter of fact,the magnetic field in the inner regions of several young stars has recentlybeen directly detected, via the Zeeman effect (Donati et al., 2005).Compared to the present-day solar system, such a hole is well within theorbit of Mercury (0.4 AU). This is also the size of a pinhole compared tothe disk sizes, which, as mentioned above, can be as large as 1,000 AU atan early stage.

Yet this ‘‘pinhole’’ region, which harbors the magnetic ‘‘central engine’’for the accretion–ejection mechanism, may have played an important role forthe early solar system. Indeed, because it is a region of tangled, unstablemagnetic fields, and located close to the star, itself the seat of its own intensemagnetic activity (as testified by the observed flaring X-ray emission, seeSection 3.2.1.2), it is a place where any circumstellar material (gas, grain,possible small planetary bodies, etc.) will suffer a high dose of radiation,either in the form of hard photons (XUV activity), or energetic particles(accelerated in stellar flares like on the Sun, and which, while not directlyobservable, can be induced from X-rays). We shall return below (Section3.2.1.2) to this important question, which connects the early evolution ofcircumstellar disks with the history of the young solar system.

To summarize, at an age of ~106 years, the young solar system – or thesolar nebula – can be depicted as follows. At its center, a magnetized solar-like star has reached most of its final mass, ~1 Mx This star is surroundedby a rigid magnetic ‘‘cavity’’ (its magnetosphere), about 0.1 AU in radius.At the ‘‘X-point’’, corotating with the star, lies the inner edge of the cir-cumstellar disk, from which most (0.9–2/3) of the matter continues to fallon the star. The remainder is ejected on both sides perpendicular to thedisk, in the form of a powerful, highly collimated jet, which evacuatesmost, if not all, its angular momentum. Beyond the X-point lies a densecircumstellar disk, either ‘‘small’’ (~50 AU, say) if it has been truncated inthe vicinity of hot stars (as in a bright, UV-rich Orion-like environment),or ‘‘large’’, a few hundred AU in radius if it was born in a dark, quietOphiuchus-like environment. Within this disk, matter continues to flowfrom the outside to the inside, braked by viscosity, until the disk isexhausted – or because matter starts to assemble to form large grains and,ultimately, planetesimals and giant planets, as discussed in the next sectionbelow.

In this ‘‘dynamic’’ picture of the solar nebula, any ‘‘heavy’’ particle like agrain can have two fates when it arrives in the vicinity of the X-point: either itfalls onto the star by accretion, or it is entrained outwards by ‘‘light’’ par-ticles (the gas), but eventually falls far back onto the disk in a ballistic fashionbecause it is too heavy to be carried away along the open magnetic field linesby the centrifugal force. In this last case, it will have spent some time veryclose to the X-ring, and will have suffered there heavy irradiation by hard


photons and energetic particles. These particles will then mix to the disk,holding specific ‘‘scars’’ from their passage near the X-ring. As explained inSection 3.2.2, this is how some models explain the mysterious presence of‘‘extinct radioactivities’’ in meteorites.

3.2. The First 10 Million Years: the ‘‘Disk Era’’

3.2.1. THE EVOLUTION OF CIRCUMSTELLAR DISKS AROUND YOUNG STARS AND

IMPLICATIONS FOR THE EARLY SOLAR SYSTEM

3.2.1.1. The path to planets: Astronomical timescale for the growth of dustgrainsJEAN-CHARLES AUGEREAU

Now that a dense circumstellar disk is installed around the central star, itmust evolve: on the one hand, it continues (albeit at a lesser rate) to lose massat its inner edge (by way of magnetic accretion), on the other hand, grainsassemble via low relative velocity collisions to form larger, preplanetarybodies. But how long does it last? Infrared observations of T Tauri stars,which are sensitive to the presence of circumstellar material, show that disksdisappear on widely different timescales. Figure 3.11 (Hillenbrand, 2006),which collects data from ~30 star-forming regions, shows that so-called‘‘inner disks’’ (i.e., regions warm enough to radiate are near-IR wavelengths)are ubiquitous at young ages, and tend statistically to disappear after a fewmillion years only. Quantitatively, the fraction is consistent with 100% at anage 1 Myr, and drops to less than 10% after 10 Myr, with some clusterscontaining no disk at all. Actually, this low fraction of ‘‘old’’ disks is a lowerlimit: mid-IR observations, which are sensitive to cooler disks, hence moredistant from the central star, show that in some cases, like the nearby g Chacluster, aged 9 Myr, the fraction of disks is closer to 40–60% (resp. Megeathet al., 2005; Lyo et al. 2003), suggesting that disks may live longer thanpreviously thought. However, at this stage the disk mass is found to be toolow to form even a Jupiter (10)3 Mx)—or perhaps they have already doneso—so that the general conclusion is that giant planets, if any, must haveformed on timescales significantly shorter than 10 Myr.

Therefore, the disk era is a critical period for planet formation, tightlyconstrained by astronomical observations. Submicron-sized dust particlescomposing young and massive disks constitute the raw material fromwhich planets form. Tiny dust grains must coagulate to form large dustaggregates, pebbles, and then larger rocky bodies (planetesimals) before thedust disk becomes too tenuous. The formation of giant planets through thecore-accretion scenario also requires the formation of planetary cores


before the disk has been mostly depleted of gas. A detailed investigation ofhow dust grains grow into large planetary embryos is presently one of themost important open questions, especially for the formation of the solarsystem: this is discussed in detail in Section 3.2.4. Here we give a broadoutline of the results drawn from the study of circumstellar disks aroundyoung stars.

A key conclusion is that the planet formation process is observationallyrequired to be both fast and common. The disappearance of circumstellardisks in less than 5–10 Myr is actually interpreted as a direct consequence ofthe formation of larger solid bodies decreasing the opacity and the dustemission (Haisch et al., 2001; Carpenter et al., 2005; Hillenbrand, 2006).Unless the circumsolar disk survived dissipation processes longer than usu-ally observed for circumstellar disks, large solid bodies in the solar systemshould have then formed within less than a few million years, which is amajor challenge for terrestrial planet formation theories.

Figure 3.11. Disk fraction in young stellar clusters, as a function of their age. This fraction isconsistent with 100% at young ages (less than 1 Myr), then declines over a timescale of a fewMyr. After 10 Myr, with a few exceptions, all the disks around young stars have disappeared,

presumably because of planet formation (Hillenbrand, 2006). This puts strong constraints onthe formation of giant planets (which cannot be seen themselves).


In addition, the growth of solid particles in disks must be sufficientlygeneric for at least two reasons. First, a significant fraction of solar-like starshave been found during the last decade to host giant gaseous exoplanets6.The formation of our solar system may then not be peculiar, and the generalconclusions derived from statistical analysis of star-forming regions may alsoapply to the early solar system. Nevertheless, it is not known whether thesegiant exoplanets formed through the accretion of gas onto a solid core or not.But there exists an independent evidence for planetesimal formation to occurroutinely in disks around young stars. A large fraction of main sequence starsin the solar neighborhood are observed to be surrounded by tenuous diskscomposed of short-lived dust grains. The survival of these so-called ‘‘debrisdisks’’, which contain much less mass that young disks (typically 10)6 Mx),over hundreds of million years points indirectly towards the presence ofreservoirs of meter-sized (or larger) bodies, similar to the Kuiper Belt in oursolar system, which release grains by mutual collisions. (We see this phe-nomenon continuing in the present-day solar system in the form of zodiacallight.) The observation of debris disks is up to now one of the most con-vincing observational clues indicating that planetesimal formation in youngdisks is common.

Various processes contribute to the coagulation and growth of dust grainsin disks of sufficiently high density. Current theoretical models of circum-stellar disks (Section 3.2.4 and references therein) include Brownian motion,vertical settling, radial drift and turbulence. Interstellar like grains (smallerthan about 0.1 micron) stay well mixed with the gas where the gas density ishigh enough. In that case, the low-velocity collisions between grains due toBrownian motion result in aggregates with fluffy structures. This process isparticularly efficient in the disk mid-plane while collisions at the disk surfacecould be destructive. As aggregates grow and reach millimeter sizes, theysettle to the disk mid-plane on short time-scales depending somewhat on thestrength of the turbulence.

Can we test this model? The available tools to witness grain growth indisks remain limited. Observations can only probe the continuum, quasi-blackbody emission of grains with sizes of the order of the observingwavelength. Therefore, astronomers are basically limited to the directdetection of solid particles smaller than, at most, a few centimeters. Fortu-nately, disks become optically thin at millimeter wavelengths, which allowsastronomers to probe the denser regions of disks where sand-like grains areexpected to reside. At these wavelengths, the total disk flux directly relies onthe dust opacity that is a function of the grain size, and it can be shown thatgrains several orders of magnitude larger than those found in the interstellar

6 Visit The Extrasolar Planets Encyclopaedia: http://www.exoplanets.eu


medium are required to explain the (sub-)millimeter observations (Draine,2006). At shorter wavelengths, the disk mid-plane is opaque and only theupper layers, mostly consisting of small grains according to the models, areaccessible. But as the opacity of disks inversely depends on the wavelength,observations should probe deeper and deeper regions as the wavelengthsincreases (Duchene et al., 2004). In practice, the interpretation of scatteredlight observations of disks at visible and near-infrared wavelengths is notstraightforward as it strongly depends on the light scattering properties of theindividual grains that are, unfortunately, hard to model properly. This typeof study is also limited to a handful of objects, such as the circumbinary diskof GG Tau (Figure 3.12)

More statistically meaningful conclusions can be derived from the study ofdust mineralogy through infrared spectroscopy. Especially, silicates that arecomposed of silicon, oxygen and, in most cases, a cation (often iron ormagnesium), have characteristic solid-state vibrational bands that occuruniquely at infrared wavelengths. The emission spectra of silicates fromoptically thin disk surfaces display characteristic emission features around 10microns due to the Si–O stretching mode, and around 20 microns due to theO–Si–O bending modes. Silicates are observed in the interstellar medium andin many solar system objects, including the Earth mantle where it appears

Figure 3.12. Left panel: observations of the dust ring about the GG Tau young binary system.In order to properly interpret the almost wavelength-independent appearance of the ring, dustsettling toward the disk midplane must be taking into account in the models (middle and right

panels). Such images may be unveiling the vertical stratified structure of the disk (Ducheneet al., 2004). The cold gas component of the GG Tau ‘‘ring world’’ has also been observed inthe mm domain (Guilloteau et al., 1999).


mostly in the form of olivine (after its olive-green color). Their presence indisks around young solar-like stars is thus expected, by analogy with thesolar system. But it is only recently that large fractions of T Tauri stars couldbe spectrally studied thanks to highly sensitive infrared space telescopes suchas Spitzer (Figure 3.13). Silicates turn out to be ubiquitous in T Tauri disks inMyr-old forming regions. But, interestingly, the strength and shape of theiremission features differ in many cases from those observed in the interstellarmedium, indicating significant processing of silicates in young disks. As thesilicate features strongly depend on grain size, the presence in disk atmo-spheres of dust particles several orders of magnitude larger than interstellargrains provides a natural explanation to the observations. Silicate emissionfeatures can thus be used as extremely valuable diagnostics of micron-sizedsolid particles in disks, as demonstrated for instance by Kessler-Silacci et al.(2006). Moreover, as the stars are fairly young (less than a few million years),this indicates fast grain growth in disks.

Figure 3.13. Ten micron silicate features from dust disks around stars of various masses

(Natta et al. 2006). From left to right: Herbig Ae stars (HAe), a few times more massive thanthe Sun), T Tauri stars (TTS), and Brown Dwarfs (BDs). Examples of theoretical silicateemission features, based on their experimental optical properties, are dispayed in the extreme

right panel (Lab.Sil.). By comparing the shape and strength of these profiles to the observa-tions, one can infer the presence of large grains as well as the degree of crystallinity of thesilicates.


The detection in circumstellar disks of crystalline silicates similar to thoseobserved in solar-system cometary dust is an additional diagnostic of similarsignificant grain processing within the first 10 Myr. The silicates that areinjected into the primordial stellar nebulae are in the form of amorphoussilicates, as proven by the recurrent non-detections of crystalline silicates inthe interstellar medium. The annealing (crystallization) of amorphous sili-cates can only happen in the inner regions of the disks where the temperatureis sufficiently high. Also their incorporation in long-orbital period solarsystem objects points towards radial (and vertical) mixing of dust during theinfancy of the solar system. Various mechanisms throughout the disk such asstellar/disk winds or radial and vertical mixing in (magneto-)turbulent diskscan, in principle, transport dust grains with important implications on thedisk chemistry and mineralogy. The observation of disks around young so-lar-like stars supports the radial mixing scenario since crystalline silicates arefound at distances where the temperature is too low for in situ annealing. Incontrast, it has been suggested that supernova shock waves in the outer solarnebula could have rapidly annealed the amorphous silicates in situ prior totheir incorporation into comets, thereby eliminating the need for large-scalenebular transport processes (Harker and Desch, 2002; see also below, Section3.2.2.3). But it is currently impossible, even for the solar system, to distin-guish between the two proposed scenarios.

It is particularly interesting to note that the statistical analysis of largesamples of T Tauri stars shows a poor correlation between the grain size inthe upper layers of disks and the age of the star. The same conclusion can bedrawn when the degree of crystallinity of the silicate grains is considered.Although star ages are difficult to estimate accurately,7 these results tend toindicate that age is not the only parameter controlling disk evolution. Theenvironment (ambient radiation field, bound companions, etc.) is likely toimpact the disk evolution and hence the planet formation which couldexplain star-to-star variations at similar ages. Although the formation ofmeter-sized (or larger) rocks is a prerequisite to form terrestrial planets andplanetary cores, one should keep in mind that this step is one of the lessobservationally nor theoretically constrained in the planet formation process.Exoplanet embryos are nevertheless expected to leave characteristic imprintsin disks such as gaps or even to clean up the inner disk regions. The detectionof gaps in young systems due to planet-disk coupling is still pending. In a fewcases, the presence of depleted regions inside disks can be indirectly inferredfrom the lack of significant circumstellar emission at short IR wavelengthsindicating a low amount of warm material close to the star. But such objectsremain exceptional. Thus, at present, astronomers are only able to draft

7 See Chapter 2


general trends without being able yet to further constrain the main steps thatallow to go from sub-micron size grain to km-sized bodies, and this is a majorproblem in particular for theories of the formation of the solar system(Section 3.2.4).

3.2.1.2. X-ray induced irradiation phenomena in circumstellar disksTHIERRY MONTMERLE

A new factor probably plays an important role, for star formation as well asfor planet formation. This is the ubiquitous X-ray emission from young stars,from the protostar stage (see, e.g., Feigelson and Montmerle, 1999; Micelaand Favata, 2005, for reviews). X-rays are mainly detected in the form ofpowerful flares, very similar to those seen in the images of the Sun which theYohkoh satellite sent us every day from 1991 to 2001,8 but much more intense(103–105 times stronger than for the quiet Sun).

In the course of these flares, the stellar X-ray luminosity LX amounts to10)4 to 10)3 times the total luminosity of the young star (called the ‘‘bolo-metric’’ luminosity Lbol), reaching a few percent or more in exceptional cases.It can be shown that the ‘‘plasma’’ (a very hot, ionized gas, with a temper-ature of 107 to 108 K) that emits the X-rays is confined in very large magneticloops (up to 2–3 R�, i.e., about 10 Rx for a T Tauri star). By analogy withthe Sun, it is thought that the fast heating results from so-called ‘‘recon-nection events’’, in which magnetic field lines of opposite polarities get incontact in a ‘‘short-circuit’’, and suddenly release the magnetic energy pre-viously stored in the course of various motions, for instance when magneticfootpoints are dragged by convective cells (e.g., Hayashi et al. 2000). Theplasma then cools radiatively typically in a few hours by the emission ofX-ray photons.

Thanks to X-ray satellites (from the 90s), X-ray emission has been detectedfrom hundreds of T Tauri stars, either concentrated in young stellar clusterslike q Ophiuchi (e.g., Ozawa et al., 2005) or Orion (Getman et al., 2005), orelse more widely dispersed as in the Taurus-Auriga clouds (e.g., Stelzer andNeuhauser, 2001), and many other star-forming regions. For a given popu-lation of T Tauri stars (with and without disks), the X-ray detection rate isconsistent with 100%. (Only the fainter stars are not all detected, for lack ofsufficient sensitivity.) In other words, observations suggest that all young,solar-like stars, emit X-rays, at a level LX/Lbol ~ 10)4 to 10)3 with a fairlylarge dispersion (see, e.g., Ozawa et al., 2005), to be compared with LX/Lbol ~ 10)7 for the present-day Sun. This is for instance most vividly illus-trated by the ~1,500 sources detected in the 10-day Chandra exposure of theOrion Nebula (Getman et al., 2005; Feigelson et al., 2005) (Figure 3.14).

8 Visit http://www.solar.physics.montana.edu/YPOP/


In addition to providing us information on the ‘‘magnetic state’’ of youngstars (for instance relevant to ejection phenomena, or to the reconnectionmechanisms, as discussed above), flare X-rays are important on anotherground: they induce irradiation effects on the dense surrounding circum-stellar and interstellar material: ionization and energetic particle interactions(this last point will be discussed in the next subsection; see Glassgold et al.,2000, 2005; Feigelson, 2005).

The first effect of X-ray irradiation is the ionization of the accretion disk.Direct evidence for irradiation has been found recently in the form of afluorescence line of neutral iron at 6.4 keV in the spectra of nearly a dozen ofT Tauri stars with disks, mainly in the q Oph and Orion clusters (Favataet al., 2005; Tsujimoto et al. 2005), and also by the presence of specificmolecules in some disks (e.g., Greaves, 2005; Ilgner and Nelson, 2006, andreferences therein). Such a characteristic fluorescence line can be excited onlyin response to X-ray irradiation of the cold gas. On theoretical grounds,several authors have studied the case of an accretion disk irradiated by X-rays (e.g., Glassgold et al., 2000, 2005; Fromang et al., 2002; Matsumura andPudritz, 2006). The results depends on details of the adopted disk model, butthe main conclusion, sketched in Figure 3.15, is that the ionization of theaccretion disk is dominated by X-rays, except in the densest equatorialregions where they cannot penetrate (‘‘dead zone’’), i.e., within a few AU ofthe central source. Everywhere else, the disk is partially ionized, and theionization fraction xe = ne/nH is roughly comparable to that of the inter-stellar medium (xe ~ 10)7 or less).

The ionization state of the disk has another important consequence: thecoupling of matter with magnetic fields. Indeed, even very weakly ionized

Figure 3.14. The center of the Orion nebula cluster. Left: near-IR image (this is the same as

Figure 3.3). Right: corresponding X-ray image by Chandra (Getman et al., 2005). Note theexcellent identification between the IR and X-ray sources, demonstrating that all young starsemit X-rays at levels much higher than the Sun.


matter ‘‘sticks’’ to the magnetic field via collisions between neutral atoms andcharged atoms (ions): this process, called ambipolar diffusion, has alreadybeen mentioned above in the context of star formation in molecular clouds(Section 3.1.1). When this process is coupled with the Keplerian motion of allatoms in a disk around the central object, it gives rise to a so-called ‘‘mag-netorotational instability’’, discovered by S. Chandrasekhar and extensivelystudied by Balbus andHawley (1991). This instability is invoked to explain thestrong viscosity of accretion disks, and thus regulates accretion itself (e.g.,Fromang et al., 2004). One interesting consequence is that such strong viscouscoupling would not exist in the (neutral) dead zone, which would then un-dergo accretion only if some other, non-magnetic mechanism for an efficientviscosity were at work (which cannot be excluded). In fact, it has even beensuggested that such a ‘‘protected’’, neutral region, of size ~20 AU, would befavorable for planet formation (Glassgold et al., 2000; Matsumura andPudritz, 2006). Note, however, that the very existence of an extended deadzone has been challenged by recent numerical computations, which take intoaccount turbulent motions within the disk, and which show that the neutral,dead zone volume tends to mix with the surrounding, ionized material: in theend, (weak) ionization probably dominates everywhere in the disk (Flemingand Stone, 2003), although this study does not take grains into account, whichturn out to play an important role in the disk ionization (Ilgner and Nelson,

Figure 3.15. Sketch of the X-ray irradiation of circumstellar disks. Note in particular the

emission of a neutral X-ray fluorescence line at 6.4 keV, detected in several systems. The otherline, at 6.7 keV, is characteristic of a hot, X-ray emitting plasma, at temperatures of several106 K. Note also the dense, neutral ‘‘dead zone’’, which, according to some authors, would be

the most favorable to planet formation.


2006). Up to now, such a weak, but widespread ionization, which must alsoaffect the growth of dust grains since they would be electrically charged, hasnot been taken into account in planet formation theories.

3.2.2. THE FIRST FEW MILLION YEARS AS RECORDED BY METEORITE DATA

MARC CHAUSSIDON, MATTHIEU GOUNELLE, THIERRY MONTMERLE

We now turn to the earliest stages of the solar system: the solar nebula, andthe so-called meteoritic record. While most of the extraterrestrial flux to theEarth is in the form of micrometer size dust (�20,000 tons of the so-calledmicrometeorites par year), there are about 10 tons of meteorites of centi-meter to meter size that fall on Earth every year. Most of these meteorites arelikely to come from the asteroidal belt situated between Mars and Jupiter.Among meteorites, chondrites are primitive objects that have endured noplanetary differenciation. This is attested by:(i) their mineralogical composition which reflects accretion of compo-

nents formed at high temperature (the chondrules) and low temper-ature (the matrix), components that were never homogenizedchemically and/or isotopically through melting and metamorphism,

(ii) their bulk composition which is similar to the photospheric abundancesof the elements in the Sun (Anders and Grevesse, 1989) and is thusconsidered as primitive, i.e. reflecting that of the forming solar system,

(iii) their age : they are the oldest rocks of the solar system, with a Rb/Sr ageof �4.55 Gyr (Wasserburg, 1987). (This early pioneering result is nowupdated by more precise measurements: see below, Section 3.2.2.2.)

Other meteorites such as irons or some achondrites are the by-products ofplanetary differentiation : they have younger ages (by a few Myr) and havecompositions reflecting the interplay of silicate–metal differentiation (core-mantle formation) and of silicate–silicate differentiation (crust-mantle for-mation). Thus chondrites, which are undifferentiated, can help us unravel thephysico-chemical conditions in the solar nebula and the astrophysical contextof the Sun’s birth, while differentiated meteorites trace the formation andevolution of large planetary bodies.

3.2.2.1. Chondrite components and physical conditions in the solar nebulaMARC CHAUSSIDON, MATTHIEU GOUNELLE

Among chondrites, carbonaceous chondrites have a chemical compositionmost similar to that of the Sun, and are therefore believed to be our bestproxy for the protosolar nebula (Brearley and Jones, 1998). The mineralogy,


chemistry and isotopic composition of carbonaceous chondrites can helpdecipher the formation and history of the solar nebula. Carbonaceouschondrites are made of Calcium–Aluminium-rich Inclusions (CAIs),chondrules and matrix, chondrules representing by far the major component,70–80% in volume (see recent reviews by Zanda, 2004 for chondrules and byMacPherson and Huss, 2003, for CAIs). (Figure 3.16)

Matrix is rich in volatile elements (e.g., H2O, C,...) : it is a fine-grainedcomponent made of chondrule fragments and of minerals stable at lowtemperature. Matrix has endured extensive secondary processing in theparent-body evidenced by metamorphic transformations (solid state diffu-sion) and hydrothermalism due to fluid-rock interactions. Matrix is thereforenot very useful for pinpointing the physical conditions in the solar nebula. Atvariance, CAIs and chondrules are high temperature components which wereformed in the solar nebula and thus predate parent-body processes. CAIs arerefractory components made of Al-, Ca-, Ti-rich silicates and oxides.

Figure 3.16. Fragment of the Allende meteorite, revealing (large white area in the bottomright-hand corner) the so-called ‘‘Calcium–Aluminium-rich Inclusions’’ (CAIs), in whichevidence for short-lived, extinct radioactivities have been found.


Accordingly, they are generally considered from their mineralogy to be thefirst solids formed by condensation from the solar nebula gas at temperatureshigher than 1800 K. Chondrules are spherical objects, made of iron–magne-sium silicates (mostly olivine and pyroxene), metal and sulfides. Chondrulesgenerally contain a large fraction of glassy mesostasis, implying that they wereonce molten and were subsequently quenched at a few to perhaps 1000 K/h(Hewins, 1997). Experimental studies show that CAIs cooled at �0.1 to�10 K/h (Stolper and Paque, 1986). Both CAIs and chondrules are thoughtto have formed in the solar nebula via complex high-temperature processesincluding condensation, evaporation, melting, etc.... Despite extensive studies,the exact mechanisms that led to the formation of CAIs and chondrules arestill elusive, but it is clear that CAIs and chondrules formed at different timesor locations. Chondrules probably formed at a higher pressure than CAIs andit is generally considered that chondrules formed later than CAIs, althoughthis assumption cannot yet be demonstrated by direct dating of each com-ponent. In fact, the formation of chondrules and CAIs may overlap and thereexists a variety of chondrules and CAIs which may have formed in differentenvironments and may have complicated histories, involving precursors ofvariable composition and more or less extensive exchanges with the nebulargas (see Hewins et al., 2005, and references therein).

3.2.2.2. Duration of the solar nebulaMARC CHAUSSIDON, MATTHIEU GOUNELLE

Radioactive nuclide abundances in rocks are classically used to date rocksand to infer timescales for geological processes.9 Among these, the U/Pbsystem which combines two parent/daughter couples (238U decays to 206Pbwith a half life of 4.47 Gyr, and 235U decays to 207Pb with a half life of0.7 Gyr), is the one which has provided the most accurate dating of mete-oritic components. Accordingly, the CAIs appear to be the oldest componentwithin chondrites, with a U/Pb age of 4567.2±0.6 Myr (Manhes et al.,1988; Allegre et al., 1995; Amelin et al., 2002). Slightly different values, basedon various methods, have appeared in the recent literature, giving an age of4568.5±0.4 Myr (Baker, 2005; Bouvier et al., 2005), but for the purpose ofthis article these values (which differ by ~1.3±0.5 Myr) are consistent withinthe uncertainties. This age can be considered as giving the ‘‘time zero’’, t0, forthe start of the formation of the solar system; however this current experi-mental time resolution of ~1 Myr does not allow to build a precise chro-nology of the earliest processes which occurred in the solar accretion disk andwhich gave birth to the first solar system solids from the nebula. In fact, the

9 See Chapter 2 on Chronometers (Section 2.2).


typical duration of these processes is most likely less than 1 Myr, so revealingthem requires further improvements in the experimental methods.

Of particular interest for cosmochemistry are the so-called extinct radio-active nuclides, or short-lived radioactive nuclides, which have half-lives be-low a few Myr. The ones which have been identified in meteorites up to noware 7Be (T1/2 = 53 days), 41Ca (T1/2 = 0.1 Myr), 36Cl (T1/2 = 0.3 Myr),26Al (T1/2 = 0.74 Myr), 10Be (T1/2 = 1.5 Myr), 60Fe (T1/2 = 1.5 Myr) and53Mn (T1/2 = 3.7 Myr). The presence of excesses of their radioactivedaughter elements (e.g., positive correlation between 26Mg/24Mg and27Al/24Mg ratios for 26Al, see Figure 3.17) shows that these radioactivenuclides were present in various amounts both in CAIs and in chondruleswhen they formed. Because of their short half-lives, short-lived radioactivitiescan be used to constrain very tightly timescales (McKeegan and Davis, 2003and refs therein): for instance the 26Al/27Al ratio decreases by a factor of twowithin 0.74 Myr. In this respect, 10Be, 26Al and 60Fe are of special interestsince they can (i) help constrain the chronology of the first million years ofevolution of the solar system, and (ii) give nuclear clues to the astrophysicalcontext of the Sun’s birth.

Many measurements of the initial content of 26Al in CAIs have led to theidea of a canonical ratio, 26Al/27Al = 4.5 · 10)5, that would have definedthe starting time t0 of the ‘‘protoplanetary’’ solar system (MacPherson et al.,1995)—which must not be confused with the starting time of the formation ofthe Sun as a star, which in this context precedes t0 (see Section 2.5). CAIs areparticularly well suited for the determination of the 26Al/27Al ratios, becausesince they are refractory objects they are enriched in Al relative to Mg. Lessnumerous measurements in chondrules, which are less enriched in Al relativeto Mg and thus more difficult to analyze, suggest that chondrules formedwith 26Al/27Al < 1 · 10)5 (Mostefaoui et al., 2002). Assuming an homoge-neous distribution of 26Al in the solar nebula, this yields a ~2–3 Myr agedifference between CAIs and chondrules. This timescale has long been con-sidered as the characteristic timescale of the solar nebula as determined bychondrite data. If true, such a long duration for the high temperature pro-cesses in the accretion disk implies that CAIs and chondrules must have beenin some way ‘‘stored’’ in the nebula for 2–3 Myr before being accreted to-gether to form the chondrites.

At this point it is important to stress that this so-called ‘‘chronologicalinterpretation’’ of the 26Al/27Al variations relies entirely on the assumptionmentioned above, which is very strong: the existence once in the early solarsystem of an homogeneous distribution of 26Al (Gounelle and Russell, 2005).This assumption is far from being proven and very difficult to demonstrate infact, simply because one would need very precise independent absolute agesfor different objects to be able to test the homogeneity of their 26Al/27Alratios. It could in fact well be that the difference in the initial content of 26Al


between chondrules and CAIs is due in part to spatial heterogeneities in thesolar nebula. Predictions can be made on the existence of such an hetero-geneity depending on the nucleosynthetic origin of 26Al, either ‘‘last minute’’injection from a nearby massive star or supernova, or production by irra-diation processes around the early Sun (see next subsection). Note also that

Figure 3.17. Variations in CAIs and chondrules of the Mg isotopic compositions (given as26Mg/24Mg in the upper panel, or as the permil 26Mg excesses noted d26Mg* in the bottom

panel) due to the radioactive decay of short-lived 26Al (which decays to 26Mg with a half-life of0.7 Myr). These data show that 26Al was present in the early solar system and that itsabundance (given as 26Al/27Al ratios) can be used to constrain the chronology of the for-mation of CAIs and chondrules. A canonical 26Al/27Al ratio of 4.5 · 10)5 has been found by

in situ ion microprobe analysis in most CAIs (data from Podosek et al., 1991, in the upperpanel). Supracanonical 26Al/27Al ratios recently found in CAIs are shown by a grey field (datafrom Young et al., 2005 and Bizzarro et al., 2004) in the bottom panel. In this panel, the

details of the distribution of 26Mg excesses in chondrules (data from Galy et al., 2002; Bizzarroet al., 2004; Chaussidon et al., 2006) is not yet well understood : it can be interpreted asreflecting either the formation of some chondrules very early (i.e. at the same time than CAIs)

or a late formation of chondrules in the nebula from a mixture of precursors including CAImaterial.


the picture has recently been complicated by results from the Oxford-UCLAlaboratory (Galy et al., 2002; Young et al., 2005) who showed that thecanonical ratio of CAIs might be due to a resetting event of an initially higherratio of �7 · 10)5, and that chondrules might have formed with an initial26Al/27Al ratio significantly higher than 1 · 10)5. The high 26Al/27Al ratiosinferred for some chondrules can indicate that they formed very early (Biz-zarro et al., 2004) or be simply due to the fact that they contain an inheritedCAI component (Galy et al., 2002).

Recently, high precision data on Pb isotopes (U/Pb dating system) havebeen obtained on CAIs and chondrules (Amelin et al., 2002), providingabsolute ages of the chondrites’ components. CAIs have ages ranging from4567.4±1.1 Myr to 4567.17±0.70 Myr, while chondrules have ages rang-ing from 4564.66±0.3 Myr to 4566.7±1 Myr. This confirms that the solarnebula could have lasted for several million years, and question the idea thatall chondrules formed 2–3 Myr after CAIs. Most recent Pb data seem toindicate that a range of ages do exist for chondrules, from nearly as early asCAIs to a few Myr later.

3.2.2.3. Extinct radioactivities and the astrophysical context of the birthof the SunMARC CHAUSSIDON, THIERRY MONTMERLE

The presence of short-lived radioactivities in the solar nebula is intriguing.Because their half-life is shorter than the timescales needed to isolate the futuresolar system material from the interstellar medium, they need to have a ‘‘lastminute origin’’, as opposed to the steady-state abundance of galactic cosmic-ray induced isotopes. They could have been injected in the nascent solarsystem by a nearby supernova, or made by irradiation of the nebular dust and/or gas by energetic solar protons, or both. The challenge of any mechanism isto explain most, if not all, of the extinct radioactivities at the same time.

The ubiquitous, intense and flare-like X-ray activity of young stars (Sec-tion 3.2.1.2) support the idea that irradiation could have been an importantprocess in the early solar system (see previous sections; also Chaussidon andGounelle, 2006, for a review of the traces of early solar system irradiation inmeteoritic components, and Feigelson, 2005 for the relation with X-ray flaresfrom young stars). Such an origin is supported by the presence of 10Be(T1/2 = 1.5 Myr) and of 7Be (T1/2 = 53 days) in Allende CAIs (McKeeganet al., 2000; Chaussidon et al., 2006), beryllium isotopes being made by flare-induced energetic particle irradiation (see Figure 3.18).

On the other hand, the presence of 60Fe in sulfides and silicates fromchondrites (Huss and Tachibana, 2004; Mostefaoui et al., 2004) implies thepresence of a nearby supernova, since 60Fe is too neutron-rich to have beenmade by in-flight spallation reactions (Lee et al., 1998), and is a signature of


explosive nucleosynthesis in supernovae. The presence of a nearby supernovawould imply in turn that the Sun was born in a crowded stellar environment,possibly similar to HII regions as observed in Orion (Hester and Desh, 2005,and above, Section 3.1.1). 26Al also can be made in massive stars andsupernova explosions (though other stellar processes, such as in novae andred giant stars, which are old, are also important sources, see Diehl et al.2006), and it is thus tempting to explain the abundance of both isotopes inthis way. As a matter of fact, one can find a set of parameters (time intervalbetween the explosion and the birth of the solar system Dt = 1 Myr, dilutionfactor due to transport of supernova material to the presolar core f=10)4)that explains both abundances, and also other isotopes (Busso et al., 2003;Gallino et al., 2004). But this may not be completely realistic: as a rule, moststar-forming regions are in some vicinity of OB associations in which massivestars evolve and explode sequentially. If they explode at a frequency shorterthan ~1 Myr, the resulting 60Fe and 26Al nucleosynthesis will tend to be in asteady state locally (as long as the massive star formation episode lasts),rather than in separate bursts, subsequently decaying. For instance, in spite

Figure 3.18. Variations in one Allende CAI of the 10B/11B ratios vs. the 9Be/11B ratio (datafrom Chaussidon et al., 2006). The isochrone-type correlation demonstrates that short-lived10Be (which decays to 10B with T1/2=1.5 Myr) was present in the early solar system. The sameCAI contains traces of the in situ decay of short-lived 7Be (T1/2=53 days). Radioactive 10Be

and 7Be can be produced in the early solar system by irradiation processes around the YoungSun (Chaussidon and Gounelle, 2006).


of the fact that the Orion Nebula Cluster currently contains no supernova,the 1.809 MeV gamma-rays from the decay of 26Al have been detected by theGRO satellite, as a signature of past supernovae from massive stars exca-vating the nearby so-called Eridanus superbubble (Diehl et al., 2004). Presentsupernovae, on the other hand, have been found by the RHESSI andINTEGRAL satellites, by way of the 1.173 and 1.333 MeV emission of 60Fein the Cygnus region, which hosts the most massive stars in the Galaxy(Harris et al. 2005). On the other hand, for the presolar core the values of Dtand f are loosely constrained since some arbitrary mass cut in the supernovaejecta has to be invoked for the calculated abundances to be in agreementwith observations in meteorites (Meyer, 2005), and neglects the precedingcontribution of the precursor massive stars. It is thus troubling that theirradiation model, which takes into account the enhanced stellar energeticparticle flux deduced from X-ray observations, should be able to account forthe 26Al/27Al ratio in CAIs independently from the possible existence of asupernova.

At this point, it should be noted that, as early as 25 years ago, Montmerle(1979) identified about 30 massive star-forming regions which were, based onvarious observational criteria, tentatively associated with supernova (SN)remnants. These ‘‘special’’ regions, dubbed ‘‘SNOBs’’ (for ‘‘OB associations’’or molecular clouds observed to be associated with supernovae) were at thetime searched in connection with the identification of high-energy gamma-raysources. For our purpose, this sample can be taken as examples of the realityof supernovae exploding in the close vicinity of young stars. It also showsthat only a small fraction of all OB associations are in this situation at anygiven time. An illustration of the complexity of the problem is given, again,by the Orion star-forming region.

In 1895, E.E. Barnard discovered a faint, almost exactly (half-) circularring in the outskirts of Orion, spanning several degrees in the sky. Thisspectacular structure, now known as Barnard’s Loop, is shown in Fig-ure 3.19. It is readily visible in the Ha line of ionized hydrogen. Its exactnature is still uncertain: proposed to be a supernova remnant by Opik back in1935, recent radio observations (Heiles et al., 2000) have shown that theemission is thermal, i.e., radiated only by an ionized gas, just like HII regionsaround young stars. The idea then is that this ring has been created by windsfrom the Belt stars (not the Trapezium; the Belt stars are the conspicuousstars visible with the naked eye on clear winter nights, see Figure 3.19). But itis also possible that the traditional radio signature of supernova remnants,i.e., the nonthermal emission from high-energy electrons accelerated at theshock front, is somehow buried in a more intense thermal emission – aclassical dilemma in supernova remnant identifications. In either case,however, kinematics indicate that Barnard’s Loop must have originatedabout 3 · 106 years ago close to the Orion nebula.


When considering the consequences on the possible irradiation of the‘‘proplyds’’ of the Orion nebula (see Figures 3.3 and 3.4), one is thereforefaced with two possibilities: (i) either Barnard’s Loop is not the remnant of asupernova explosion having taken place 3 · 106 years ago, in which casenone of the current young stars in Orion have been peppered with 60Fe, (ii) orit is, then only a fraction of them have been. This can be seen by looking atthe H–R diagram presented in Chapter 2 on chronometers (Figure 2.3). Inthis figure, we can easily see that the majority of young stars present in theOrion Nebula Cluster are in fact younger that 3 · 106 years, so that any 60Fespread by the explosion has disappeared, and these young stars cannot becontaminated by 60Fe. Quantitatively, one can find after Figure 3.20 (takenfrom Palla and Stahler, 1999), that less than 40% of the older generation ofyoung ‘‘suns’’ may have been effectively contaminated by the SN explosion, if

Figure 3.19. Barnard’s Loop surrounding the Orion nebula, seen in Ha. It is unclear whetherthis extended structure is a the remnant of a supernova explosion or an ionized shell created bystellar winds from the Belt stars, but kinematical studies give it an age of 3 · 106 years.(Photograph by E. Mallart).


there was such an explosion. The presence of 60Fe in the early solar systemwould then not be the rule, even in an Orion-like birthplace.

In the end, and as discussed in the introductory sections, the birthplace ofthe Sun is still an unresolved question, although the birth of the Sun in a richcluster seems to be favored by stellar statistics. To understand whether the Sunwas born in a high-mass environment like Orion, or in a low-mass environ-ment like q Ophiuchi or Taurus has however important implications not onlyfor the astrophysical conditions for the Sun’s birth itself, but also for thechronology of the early solar system. Indeed, depending on their origin, short-lived radionuclides were or were not homogeneously distributed in the solarnebula. Usually short-lived radionuclides are expected to be homogeneouslydistributed if coming from a supernova and heterogeneously distributed iforiginating from in situ irradiation by energetic particles. This comes from thefact that supernova material must be largely volatilized in the HII region andhomogeneously mixed in the accreting disk where it is injected, though there isat present no definitive observation demonstrating whether a fraction of the

Figure 3.20. Age distribution in four mass ranges for stars in the Orion Nebula Cluster (Pallaand Stahler, 1999). The dotted line indicates the estimated age of Barnard’s Loop, 3 Myr. Forstars of mass close to 1 Mx, less than 40% are older than 3 Myr and may have been

contaminated by short-lived nucleosynthetic products of a supernova explosion, providedBarnard’s Loop is a supernova remnant, which is still unclear.


supernova ejecta cannot be in the form of solid grains (Hester and Desch,2005). At variance, irradiation models based on X-ray flare observations andthe ‘‘X-ring’’ picture (Gounelle et al., 2001; 2004; see above, Section 1.3)predict possible variations in the production rate of short-lived radionuclidesdepending on parameters such as the fluence of the accelerated particles, theircomposition and the composition of the irradiated target. Irradiation, how-ever, might also produce rather constant radionuclide abundances if charac-teristic time scales and compositions are considered for the irradiation. It isobvious that identifying the source of 26Al and other short lived nuclideswould provide a long-needed basis for the chronological use of these elements.

One could think of solving this problem by looking at young stars withcircumstellar disks, in which particle irradiation is actually taking place asseen in X-rays, making use of the fact that, as mentioned above, 26Al decaysby the emission of 1.8 MeV gamma-rays, and is thus observable elsewherethan in the solar system by gamma-ray telescopes such as GRO. The cal-culation has been done by Montmerle (2002): taking into account thatgamma-ray telescopes have a very wide field-of-view (several degrees indiameter), when pointed at a star-forming region they will integrate the fluxof a whole star-forming region, i.e., several hundred young stars at the sametime. As it turns out, even under the most optimistic assumptions the1.8 MeV flux is undetectable, and dominated by the general 26Al emission inthe Galaxy, which is most conspicuous in massive star-forming regions be-cause of successive supernova explosions (see Diehl et al. 2006).

In summary, while the presence of 60Fe in CAIs shows that the formingsolar system has been at least ‘‘polluted’’ by a nearby supernova, it is not clearwhether this supernova has been responsible for the other extinct radioac-tivities, including 26Al.

3.2.3. INTERMEZZO

THIERRY MONTMERLE

The study of circumstellar disks around solar-like stars, as described inSection 3.1.2, obviously has strong implications on our current views on theorigin and formation of the solar system. We observe their global evolution(timescale for the disappearance of disks in stellar clusters), and we begin tounderstand their physicochemical evolution (growth and mineralogy of dustgrains via IR spectroscopy) over a few million years, while large quantities ofgas (detectable via mm observations) are still present. However, as alreadymentioned, once the grains reach sizes of a few mm, they are not observableany more. They start to be observable again much later (several tens ofmillion years), but only as ‘‘debris’’ of collisions between macroscopic bodies,presumably planetesimals or asteroids, giving indirect evidence for ongoing


planetary formation. Thus, there is a crucial observational time gap betweenthe ‘‘primordial’’ disks (which would correspond to the solar nebula), and‘‘second-generation’’ debris disks (in which planetary formation is well underway). Translated into sizes, this means that we have essentially no observa-tional constraints on the transition between cm-sized grains (direct) andkm-sized bodies (indirect).

In principle, the ~200 known exoplanetary systems should also give usclues about planetary formation in general, and the formation of the solarsystem in particular. However, at least as far as the formation of the solarsystem is concerned (which is our main concern here in the context of theorigin of life as we know it), there are still many open problems. Indeed, whilecurrent observing techniques are sensitive mostly to super-giant exoplanetsclose to the central stars (the so-called ‘‘hot Jupiters’’, with masses between 5and 10–15 Jupiter masses or more), and increasingly more to lower-massexoplanets, exoplanetary systems with several planets, when they exist, donot resemble at all the solar system. The recently announced discovery of adistant 5-Earth-mass exoplanet orbiting a very low-mass, non-solar, star, byway of gravitational lensing towards the galactic center (Beaulieu et al. 2006),is not helpful either in this context, since nothing is known about the possibleexoplanetary system it belongs to.

In the end, for most of the phase of planet formation we are essentially leftwith theory, backed by rare and difficult laboratory experiments. The goal isto put together mechanisms for grain agglomeration and destruction, andmutual dynamical interactions between bodies of various sizes and masses inthe general gravitational field of the central star, and also in the presence ofgas which affects the motion of small particles. In spite of the relative sim-plicity of the basic equations, this involves sophisticated numerical N-bodysimulations. The following section describes the basic theoretical concepts ofplanetary formation restricted to the solar system, with the goal of ultimatelyunderstanding the various steps that led to the formation of our planet, theEarth.

3.2.4. THE FIRST STAGES OF PLANETARY FORMATION IN THE SOLAR SYSTEM

ALESSANDRO MORBIDELLI

3.2.4.1. From micron-size to kilometer-size bodies (1 Myr)In the proto-solar nebula, the most refractory materials condense first,gradually followed – while the local temperature drops – by more and morevolatile elements. The dust grains thus formed float in the gas. Collisionsstick the grains together, forming fractal aggregates, possibly helpedby electrostatic and magnetic forces. Other collisions then rearrange the


aggregates, and compact them. When the grains reach a size of about acentimeter, they begin to rapidly sediment onto the median plane of the diskin a time

Tsed � R=ðqpXaÞ � ðqvÞ=ðqpX2aÞ

where R is the nebula surface density, q is its volume density, v is the r.m.s.thermal excitation velocity of gas molecules, qp the volume density of theparticles, a is the radius, and X is the local orbital speed of the gas (Goldreichand Ward, 1973; Weidenschilling, 1980). Assuming Hayashi’s (1981) minimalnebula (minimal mass solar composition nebula, with surface density Rproportional to r)3/2, containing materials to create the planets as we knowthem), one gets

Tsed � 103=a years

where a is given in cm. This timescale, however, is computed assuming aquiet, laminar nebula. If the nebula is strongly turbulent, or strongly per-turbed from the outside or by the ejection of jets in the proximity of the star,the sedimentation can become much longer.

Once the dust has sedimented on the mid-plane of the nebula, the clockmeasuring the timescale of planet accretion starts effectively to tick. Thus, thetime t0 for planetary accretion is not the time t0 usually used in stellar for-mation theory (start of the collapse of molecular cores) or the time t0 ofcosmochemists (the formation of the first CAIs, see above, Section 3.2.2.2).Linking the various times t0 together is one of the major problems inestablishing an absolute chronology for the formation of the solar system. Inaddition, notice that Tsed above depends on the heliocentric distance r. Thismeans that time t0 is different from place to place in the disk!

The growth from dust grains to kilometer-size planetesimals is stillunexplained. There are two serious issues that remain unsolved. The firstissue concerns the physics of collision between such bodies at speed of order10 m/s (typical collision velocity for Keplerian orbits with eccentricitye ~ 10)3), which is still poorly understood. For dust grains, current theoriespredict that collisions are disruptive at such speeds (Chokshi et al., 1993).However, recent laboratory experiments on collisions between micrometersize grains (Poppe and Blum, 1997; Poppe et al., 2000) give a critical velocityfor accretion (velocity below which grains accrete, and above which theyfragment) 10 times larger than predicted by the theory. The reason is prob-ably that the fractal structure of the dust allows to absorb energy much betterthan envisioned a priori. This may help solving the fragmentation paradoxfor dust–dust collisions. When the agglomeration of dust builds larger bod-ies, though, the problem of collisional disruption becomes much more severe.Laboratory experiments cannot be done at this size range and one has to


trust computer models. Specific simulations of this process with SPH(Smooth Particle Hydrodynamics) techniques have been done by Benz (2000)for basalt bodies (monolithic or rubble piles) of sizes in the range m to km.He found that low velocity collisions (5–40 m/s) are, for equal incomingkinetic energy per gram of target material, considerably more efficient indestroying and dispersing bodies than their high velocity counterparts.Furthermore, planetesimals modeled as rubble piles are found to be char-acterized by a disruption threshold about five times smaller than solid bodies.Thus, unless accretion can proceed avoiding collisions between bodies ofsimilar masses, the relative weakness of bodies in meter to km size rangecreates a serious bottleneck for planetesimal growth. These apparently neg-ative results, however, may once again depend on our poor understanding onthe internal structure of these primitive small planetesimals. The simulationsassume rocky objects, but it is still unclear how a puffy pile of dust becomes asolid rock. If the planetesimals were not yet ‘solid rocks’ maybe the impactenergy could be dissipated more efficiently.

The second open problem on planetesimal growth concerns radialmigration. Gas drag makes them fall onto the central star: gas being sus-tained by its own pressure, it behaves as if it felt a central star of lower mass,and therefore rotates more slowly than a purely Keplerian orbit at the sameheliocentric distance. Solid particles tend to be on Keplerian orbits andtherefore have a larger speed than the gas. So the gas exerts a force (drag) onthe particles, the importance of which is given by the characteristic stoppingtime:

Ts ¼ ðmDVÞ=ðFDÞ;

where m is the mass of the grain, DV the difference between the particlevelocity and the gas velocity, and FD the gas drag. For small particles, the‘‘Epstein’’ gas drag gives

Ts � qp=R;

while for big particles, the ‘‘Stokes’’ drag gives

Ts � qpa2:

(Weideschilling, 1977). We then compare this time with the characteristicKeplerian time

TK ¼ 1=X:

If Ts>>TK, the particle is almost decoupled from the gas. If Ts<<TK, it isstrongly coupled to the gas, and it tends to move with the gas flow. Themaximum effect occurs when Ts ~TK, which occurs for meter sizeparticles. For the Hayashi minimal nebula, at about r = 1 AU from the Sun,


DV~ 50–55 m/s, and the particle’s radial velocity induced by the gas drag isthen of order 10–100 m/s (Weideschilling, 1977). So meter-sized particlesshould fall on the Sun in ~100–1,000 years, i.e., before they can grow massiveenough to decouple from the gas.

One way out of this paradox is to have a density of solids larger than thatof the gas. In this case, the growth timescale would be faster than the radialdrift timescale. However, such a composition is not supported by observa-tions nor by current theories.

Another possibility is the existence of vortices in the protoplanetary disk,due to the turbulent viscosity of the nebula (Tanga et al., 1996). In thismodel, 70–90% of the particles are trapped in anti-cyclonic vortices. Oncetrapped, the particles do not fall any more towards the Sun, but rather falltoward the center of the vortex. Such falling timescale varies from a fewtens to a few thousand years, depending mainly on the size of the particleand on the heliocentric distance (which increases all dynamical timescales).Once at the center of the vortex, particles dynamics are stable over thelifetime of the vortex. Their relative velocities are reduced (particles tend tofollow the gas stream lines, so they all tend to have the same velocity) andthe local density is increased, enhancing the accretion process. Therefore,vortices would help the accretion of kilometer-size planetesimals in twoways: by stopping the drift of meter-sized bodies towards the Sun, and byspeeding up the accretion process due to the accumulation of the bodies atthe centers of the vortices.

Which of these two possible situations is the real one, profoundly affectsthe formation timescale. If there is no way to slow down the fall of growingplanetesimals towards the Sun, then the formation of a multi-km object hasto occur in about a few 1,000 years, probably by gravitational instability. If,on the contrary, the turbulence of the disks is an effective obstacle to theinwards drift, then the formation of planetesimals can take much longer. Toadd confusion (reality is never easy), it is likely that while the first plane-tesimals are building up, new dust is settling on the midplane or drift in fromfurther heliocentric distances. So, different planetesimals can see different‘‘times t0’’. In other words, even if the formation of a planetesimal is locallyvery fast, the formation of a population of planetesimals can be ongoing formuch longer.

3.2.4.2. Formation of planetary embryos (1–10 Myr)One way or another, we now have a gas disk containing kilometer-sizeplanetesimals. The dynamics of accretion starts to be dominated by the effectof the gravitational attraction among the planetesimals, which increases thecollisional cross-sections. A runaway growth phase starts, during which thebig bodies grow faster than the small ones, hence increasing their relative


difference in mass (Greenberg et al., 1978). This process can be summarizedby the equation:

d

dt

M1

M2

� �¼M1

M2

1

M1

dM1

dt� 1

M2

dM2

dt

� �>0;

whereM1 andM2 are, respectively, the characteristic masses of the ‘‘big’’ andof the ‘‘small’’ bodies, and can be explained as follows.

Generally speaking, accretion is favored by a high collision rate, whichoccurs when the relative velocities are large, but also by large collisionalcross-sections and gentle impacts, which occur when the relative velocities arelow. Therefore the relative velocities between the different planetesimalpopulations govern the growth regime.

At the beginning of the runaway growth phase the large planetesimalsrepresent only a small fraction of the total mass. Hence the dynamics isgoverned by the small bodies, in the sense that the relative velocities amongthe bodies is of order the escape velocity of the small bodies Vesc(2). Thisvelocity is independent of the mass M1 of the big bodies and is smaller thanthe escape velocity of the large bodies Vesc(1). For a given body, the collisionalcross-section is enhanced with respect to the geometrical cross-section by theso-called gravitational focusing factor:

Fg ¼ 1þ ðV2esc=V

2relÞ;

where Vesc is the body’s escape velocity and Vrel is the relative velocity of theother particles in its environment. Because Vrel ~Vesc(2), the gravitationalfocusing factor of the small bodies [Vsec = Vvesc(2)] is of order unity, whilethat of the large bodies [Vesc = Vesc(1)>>Vesc(2)] is much larger. In thissituation one can show that mass growth of a big body is described by theequation

1

M1

dM1

dt

� ��M

1=31 V�2rel

(Ida and Makino, 1993) Therefore, the relative growth rate is an increasingfunction of the body’s mass, which is the condition for the runaway growth(Figure 3.21).

The runaway growth stops when the mass of the large bodies becomesimportant (Ida and Makino, 1993) and the latter start to govern thedynamics. The condition for this to occur is:

n1M21 > n2M

22;

where n1 (resp. n2) is the number of big bodies (resp. small bodies). In thiscase, Vrel ~Vesc(1) ~M1

1/3, and hence (1/M1)(dM1dt) ~M11/3. The growing


rate of the embryos gets slower and slower as the bodies grow, and therelative differences in mass among the embryos also slowly become smaller.In principle, one could expect the small bodies themselves to grow, narrowingtheir mass difference with the embryos. But in reality, the now large relativevelocities prevent the small bodies to accrete with each other. The small

Figure 3.21. A simulation of the runaway growth process for planetary embryos. In a disk ofequal mass planetesimals, two ‘‘seeds’’ (planetesimals of slightly larger size) are embedded. As

time passes, the two seeds grow in mass much faster than the other planetesimals,, becomingplanetary embryos (the size of each dot is proportional to its mass). While the growingplanetary embryos keep quasi-circular orbits, the remaining planetesimals have their eccen-tricities (and inclinations) excited by the close encounters with the embryos. Notice also that

the separation between the embryos slowly grows in time (i.e. passing from one panel to thesubsequent one). From Kokubo and Ida (1998).


bodies can only participate to the growth of the embryos: this phase is called‘‘oligarchic growth’’.

The runaway growth phase happens throughout the disk, with timescalesthat depend on the local dynamical time (Keplerian time) and on the localdensity of available solid material. This density will also determine themaximum size of the embryos and/or planets when the runaway growth ends(Lissauer, 1987). Assuming a reasonable surface density of solid materials,the runaway growth process forms planetary embryos of Lunar to Martianmass at 1 AU in 105–106 yr, separated by a few 10)2 AU. Beyond the so-called snow line at about 4 AU, where condensation of water ice occurredbecause of the low temperature, enhancing the surface density of solidmaterial, runaway growth could produce embryos as large as several Earthmass in a few million years (Thommes et al., 2003).

3.2.4.3. Formation of the giant planets (10 Myr)Observations and models of the interior of the giant planets give someimportant constraints on the composition and mass of the giant planets (seeGuillot, 1999, and for a review, Guillot, 2005):(i) Jupiter has a mass of 314 ME (Earth mass), and contains ~10–30 ME

of heavy elements;(ii) Saturn has a mass of 94 ME and contains ~10–20 ME of heavy ele-

ments;(iii) Uranus and Neptune have a mass of 14 and 17 ME respectively, of

which only ~1–2 ME of hydrogen and helium.

To account for these constraints, the best current models for the formationof Jupiter and Saturn assume a three-stage formation (Pollack et al., 1996):(1) the solid core accretes as explained in the previous section; beyond the

snow line, the surface density of solid material is enhanced by a factor ofseveral, due to the presence of ice grains. This allows embryos to grow toabout 10 ME on a timescale of a million years (Thommes et al., 2003).

(2) The accretion of the solid core slows down (see above), while a slowaccretion of nebular gas begins, due to the gravity of the core. The gasaccretion continues at a roughly constant rate over many million years,until a total mass of 20–30 ME is reached;

(3) When the mass of the protoplanet reaches ~20–30 ME the gas gravita-tionally collapses onto the planet. The mass of the planet grows expo-nentially and reaches hundreds of Earth masses in ~10,000 years: this isthe ‘‘runaway’’ phase. (Figure 3.22)

The model explains well the properties of Jupiter and Saturn, in particularthe existence of solid cores of about 5–15 Earth masses. There are howeverfour main problems in the above scenario, which have not yet been solved.


(I) When the planetary core reaches a mass of several Earth masses, itstidal interaction with the gas disk forces it to migrate very rapidly towardsthe Sun. (This is called ‘‘Type-I migration’’). The estimated falling time ismuch shorter than the time required for the onset of the exponential accre-tion of the massive atmosphere. Thus, giant planets should not exist! Twoways out of this paradox have been proposed.

The first is that the gas disk was violently turbulent. In this case, theplanetary cores would have suffered a random walk, rather than a monotonicinfall towards the star (Nelson, 2005). Some would have collided with the stareven faster than in the absence of turbulence, but others could be luckyenough to avoid collisions for a time long enough to start phase 3 above.

A second possibility (Masset et al., 2006) is that the gas disk surfacedensity had a radial discontinuity. For instance, the inner part of the diskcould be depleted by a factor of a few by the ejection of material in the polarjets, typical of young – magnetically active – stars (Section 3.1.3). If such adiscontinuity exists, the planetary core would migrate towards the disconti-nuity, and stop there until the atmosphere is accreted.

(II) The second phase of giant planet accretion (the slow accretion of theatmosphere, prior to the onset of the runaway growth of the giant planet’smass) is also a problem, as it takes about 10–15 Myr, longer then the typical

Figure 3.22. The growth of a Jupiter-mass planet. The solid curve gives the mass of metals as afunction of time. The dotted curve gives the mass of the gas, and the dash-dotted curve the

sum of the two, as a function of time. Notice that the growth of the solid core of theplanet almost stalls after the first 0.5 Myr. During a ~7 Myr timespan, the planet slowlyacquires an atmosphere, and, only when the total mass overcomes a critical threshold, a finalexponential accretion of gas is possible. The timescale characterizing the slow accretion of the

gas depends on the opacity of the atmosphere. From Pollack et al. (1996).


nebula dissipation time (Haisch et al., 2001; Hillenbrand 2006) (see above).To shorten the timescale of the second phase, two solutions have been pro-posed: to have an enveloppe of reduced opacity (Podolack 2003), or themigration of the planet, which continuously feeds the growth of its core(Alibert et al., 2005).

(III) The end of the exponential gas accretion is not yet fully understood.Most likely the growth of the giant planets is slowed down when a gap isopened in the gaseous disk, and is finally stopped when the nebula is dissi-pated by photo-evaporation from the central star. If this is true, then the fullformation timescale of Jupiter and Saturn is of the order of the lifetime of thegas disk, namely of a few Myr. For Uranus and Neptune, it is generallyassumed that the nebula disappeared before that the third phase of accretioncould start. This would explain why these two planets accreted only a fewEarth masses of gas.

(IV) The last problem is that, at the endof stage (III) above, the giant planetsopen a gap in the gas disk. Consequently they become locked in the radialevolution of the disk. As the disk’s material tends to be accreted by the star, thegiant planets have to migrate inwards. This migration, although slower thanthat discussed above for the cores, is nevertheless quite fast. (This is the so-called ‘‘Type-II’’ migration.) It is usually invoked to explain the existence of‘‘hot Jupiters’’, massive extra-solar planets that orbit their star at distancessmaller than the orbital radius ofMercury. But in our solar system this kind ofmigration evidently did not happen, or at least did not have a comparably largeradial extent. Again, two solutions to this problem have been advanced.

The first possibility is that Jupiter and Saturn formed sufficiently late thatthe disk was already in the dissipation phase. Thus, the disk disappearedbefore it could significantly move the planets. In addition, this solution hasthe advantage of explaining why Jupiter and Saturn did not grow further inmass and why their massive atmospheres are enriched in heavy elementsrelative to the solar nebula composition (e.g., Guillot and Hueso, 2006).

A second possibility is that Jupiter and Saturn formed almost contem-poraneously and on orbits that were close to each other. In this case, the gapsopened by the two planets in the disk would have overlapped (Masset andSnellgrove, 2001). This would have changed dramatically the migrationevolution of the planets pair, possibly stopping or even reversing it. Of coursethe two solutions imply different formation timescales. If the planets stoppedbecause the gas disappeared, the fact that Jupiter and Saturn did not migratesignificantly implies that their formation timescale is of order of the nebuladissipation time (3–10 Myr; Hillenbrand 2006). In the opposite case, theymight have formed even faster.

Finally, to be complete, a model of giant planet formation alternativeto that of Pollack et al. has been proposed by Boss (see Boss, 2003 and ref-erences herein). In this model, the proto-planetary gas disk was massive


enough to be unstable under its own gravity. In this situation, gaseous planetscan form very rapidly, by a process that reproduces in miniature the one thatled to the formation of the central star. There is no need of the presence of amassive solid core to trigger the capture of the giant planet’s atmosphere. Inthis sense, Boss’s model may bring a solution to the timescale problem relatedwith the Pollack et al. model, discussed above. There is a still open debate inthe literature on whether the clumps of gas observed in numerical simulationsof gravitationally unstable disks are temporary features or would persistuntil the disk’s dissipation. For the case of our solar system, the presence ofmassive cores inside all giant planets, and the limited amount of gas in Uranusand Neptune, make us think that the Pollack et al. model is more appropriate.It is possible, however, that some or several of the extra-solar planets observedso far formed through a gravitational instability mechanism.

3.3. The First 100 Million Years: the ‘‘Telluric Era’’

3.3.1. FORMATION OF THE TERRESTRIAL PLANETS AND PRIMORDIAL

SCULPTING OF THE ASTEROID BELT


After a few 105 years of runaway growth, the embryos in the terrestrial planetregion and in the asteroid belt region have Lunar to Martian masses. Theygovern the local dynamics and start perturbing each other. The system be-comes unstable, and the embryos’ orbits begin to intersect (Chambers andWetherill, 1998). Because of mutual close encounters, the embryos’ dynam-ical excitation (increase of eccentricity and inclination) moderately increases,and accretional collisions among embryos start to occur. The situationdrastically changes when Jupiter and Saturn acquire their current masses.These two planets strongly perturb the dynamical evolution of the embryosin the asteroid belt region between ~2 and 5 AU. The embryos acquire astrong dynamical excitation, begin to cross each other, and cross ratherfrequently the orbits of the embryos in the terrestrial planets region. Thecollision rate increases. Despite the high relative velocity, these collisions leadto accretion because of the large mass of the embryos.

The typical result of this highly chaotic phase – simulated with severalnumerical N-body integrations – is the elimination of all the embryos orig-inally situated in the asteroid belt and the formation a small number ofterrestrial planets on stable orbits in the 0.5–2 AU region in a timescale~100 Myr (Figure 3.23).

This scenario has several strong points:


(i) Planets are formed on well separated and stable orbits only inside~2AU. Their number typically ranges from 2 to 4, depending on thesimulations, and their masses are in the range Mars mass – Earth mass(Chambers and Wetherill, 1998; Agnor et al., 1999).

(ii) Quasi-tangent collisions of Mars-mass embryos onto the proto-planets are quite frequent (Agnor et al., 1999). These collisions areexpected to generate a disk of ejecta around the proto-planets (Canupand Asphaug, 2001), from which a satellite is likely to accrete (Canupand Esposito, 1996). This is the standard, generally accepted, scenariofor the formation of the Moon (see below, Section 3.2)

(iii) The accretion timescale of the terrestrial planets is ~100 Myr. This iscompatible with several constraints on the chronology of accretioncoming from geochemistry (Allegre et al. 1995). On the other hand,Hf–W chronology seems to indicate that the formation of the Earth’score occurred within the first 40 Myr (Yin et al., 2002; Kleine et al.,2002). This might suggest that the Earth accretion was faster than itappears in the simulations. However, if the cores of the embryos arenot mixed with the mantles during the collisions – as indicated bySPH simulations (Canup and Asphaug, 2001) – this timescale wouldmeasure the mean differentiation age of the embryos that participated

Figure 3.23. The growth of terrestrial planets from a disk of planetary embryos. Each panelshows the semi-major axis and eccentricity of the bodies in the system, the size of each dot

being proportional to the mass. The color initially reflects the starting position of each em-bryo. When two (or more) embryos collide, the formed object assumes the color corre-sponding to the embryo population that has mostly contributed to its total mass. A system of

four terrestrial planets, closely resembling our solar system, is formed in 200 Myr. FromChambers (2001).


to the formation of the Earth, and not the time required for our planetto accrete most of its mass.

(iv) All the embryos located beyond 2 AU are eliminated in 2/3 of thesimulations (Chambers and Wetherill, 2001). They either aredynamically ejected from the solar system, or collide with the Sun, orare accreted by the forming terrestrial planets.

(v) In the same time, the small planetesimals are subject to the combinedperturbations of the giant planets and of the embryos (Petit et al.,2001). The dynamical excitation increasing very rapidly (timescale 1–2 Myr), most of the small planetesimals are eliminated in a few millionyears by either the ejection from the solar system, or the collision withthe Sun or with a growing planet. In the asteroid belt (2–4 AU range),this leads to a remaining population of small bodies (the asteroids) onstable orbits with quite large eccentricities and inclinations, whichcontains only a very small fraction of the total mass initially in theregion. This scenario explains well the current mass deficit of theasteroid belt, the eccentricity and semi-major axis distribution of thelargest asteroids and other more subtle properties of the asteroid beltpopulation, such as the partial mixing of taxonomic types.

However, this scenario of terrestrial planet formation suffers from someweaknesses:(i) The final orbits of the planets formed in the simulations are typically

too eccentric and/or inclined with respect to the real ones. This couldbe due to the fact that the current simulations neglect the so-calledphenomenon of dynamical friction, namely the effect of a large pop-ulation of small bodies, carrying cumulatively a mass comparable tothat of the proto-planets. Dynamical friction should damp theeccentricities and inclinations of the most massive bodies.

(ii) Obliquities of the terrestrial planets should have random values.However in reality, only one planet has a retrograde spin (Venus).Moreover all planetary obliquities are compatible with an initial 0-degrees obliquity, modified by the subsequent evolution in theframework of the current architecture of the planetary system (Laskarand Robutel, 1993).

(iii) The planet formed in the simulations approximately at the location ofMars is typically too massive.

3.3.2. THE FORMATION OF THE MOON


As explained above, the currently accepted model for the Lunar formation isthat of a giant impact occurring during the formation of the Earth. The


current view of terrestrial planet formation implies several giant impacts, andimpacts with an angular momentum similar of that of the Earth–Moonsystem are not rare, particularly during the end of the accretion process(Agnor et al., 1999). Simulations of a Moon-forming impact have been donesince 1986, using SPH simulations. The most advanced, recent high-resolu-tion simulations have been done by Canup and Asphaug (2001) and Canup(2004a). A very detailed review on the Moon formation can be found inCanup (2004b).

In the SPH simulations, three critaria has been used to judge the degree ofsuccess: the formation of a circumplanetary disk with about a Lunar massoutside of the Roche radius of the Earth, a mass of iron in the protolunardisk that is about 10% of the total mass (the fraction present inside theMoon), and a total angular momentum of the Earth-disk system of order ofthat of the current Earth–Moon system. In essence, in case of an ‘‘early’’formation of the Moon, when the proto-Earth was only 60% of the currentEarth mass, the impactor needs to be of about 30% of the total mass (proto-Earth + impactor). If the impact is late, the impactor can be of about 10%of the total mass, namely of order of the mass of Mars. The authors privilegethe ‘‘late impact’’ scenario, because otherwise the Moon would have accretedtoo many siderophile elements after its formation, assuming that it accretedabout 10% of the mass that is required to collide with the Earth to completethe Earth’s formation and a chondritic compostion of such material.

The differentiation of the Moon can be dated using the Hf–W chro-nometer, and turns out to have occurred at about 40 Myr after CAI for-mation (Yin et al., 2002; Kleine et al., 2002). Thus, if the Moon-formingimpact was the last one (or close to the last one), the Earth formed (i.e.,received its last giant impact) in a similar timescale. If on the contrary, theMoon-forming impact was an early one, the formation of the Earth mighthave taken longer. The Hf–W chronometer seems to indicate an age of 40 Myalso for the differentiation of the Earth, but in this case the interpretation isless straightforward because, in case there is little equilibration between thecore and the Mantle during the giant impacts, the overall mechanicalaccretion process of the Earth could have taken significantly longer (Sasakiand Abe, 2004).

In the SPH simulation of the Moon-forming impact, about 80% of thematerial that ends in the proto-lunar disk comes from the impactor, ratherthan from the proto-Earth. The Moon and the Earth have a very similarcomposition under many aspects (for instance the oxygen isotopic compo-sition is identical). The logical interpretation is then that the proto-Lunarimpactor and the proto-Earth had very similar compositions (and identicaloxygen isotope composition). This however seems in conflict with the resultsof N-body simulations of the accretion of terrestrial planets. These simula-tions show that a planet forms by accreting material in a stochastic way from


a wide variety of heliocentric distances. Given that the oxygen isotopecompostion of Mars and of all meteorite classes differ from each other, itthen seems unlikely that the proto-Earth and the lunar impactor could haveexactly the same resulting composition. A solution of this apparent paradoxhas been proposed by Pahlevan and Stevenson (2005). During the formationof the proto-lunar disk there might have been enough isotopic exchange withthe proto-Earth to equilibrate the two distributions. Another possibility isthat the fraction of the proto-Lunar disk of terrestrial origin was in realitymuch larger than estimated in the SPH simulations. In fact these simulationsassume a non-spinning Earth, so that the totality of the angular momentumof the Earth-Moon system is carried by the impactor. This has the conse-quence of having the impactor on a quasi-tangent trajectory relaitve to theproto-Earth, which maximizes the amount of impactor material that goes ingeocentric orbit. If the Earth was spinning fast at the time of the Moonformation, then the impactor might have had a more head-on trajectory andless of its material would have contaminated the proto-Lunar disk. SPHsimulations with a fast rotating Earth have never been done up to now.

The accretion of the Moon from a protolunar disk has been simulated byIda et al. (1997) and Kokubo et al. (2000). The Moon forms very quickly, inabout 1 year, at a distance from the Earth of about 1.2 to 1.3 Roche radii. Inmany cases, the Moon is accompanied by many moonlets, left-over of theaccretion process in the disk. More rarely, two Moons of similar mass areformed. The subsequent evolution of these systems, simulated in Canupet al. (1999), typically leads to an end-state with a single Moon on a stableorbit.

3.3.3. TOWARDS 1 GYR: THE EARLY EVOLUTION OF THE EARTH

BERNARD MARTY

3.3.3.1. Formation and closure of the Earth’s atmosphereThe formation of the terrestrial atmosphere is a suite of complex processeswhich are not yet fully understood. Indeed its abundance and isotopiccompositions differ from those of the solar nebula and of other knowncosmochemical components like comets or primitive meteorites. The generalconsensus is that our atmosphere was formed by contributions of severalvolatile-rich components and that it was subject to several episodes of loss tospace throughout its early history. Some of these loss events werenon-fractionating, that is, elements and isotopes were lost in the same pro-portion. This might have been the case of large impacts. Because the isotopiccomposition of some of the atmospheric elements such as noble gases differfrom known end-member compositions, one has to consider also that


isotopic fractionation took place, and that there are several escape processesfor the atmosphere that have been demonstrated to isotopically fractionatenoble gases. Among these, there are the thermal loss, in which the velocity ofa given isotope is higher than that of another given isotope, allowing theformer to escape at a larger rate, and the pick-up ion loss in which atoms atthe top of the atmosphere are ionized during charge transfer from solar windions, light isotopes being statistically more prone for such ionization. Henceit is convenient to define an epoch at which the atmosphere became closed tofurther loss into space.

During at least the first tens of Myr of its existence, while the Sun was stillon its way to the main sequence, the Earth was subject to large-scale igneous(volcanic) events, during which the proto-mantle exchanged volatile elementswith surface reservoirs. It is likely that a primitive atmosphere existed at thistime, but any chemical record of it has been erased owing to the high thermalstate of the Earth evidenced by core formation and magma ocean episodes.Records of atmospheric processes at this time cannot be found directly at theEarth’s surface at Present. The record of extinct radioactivity systems inwhich parents differ from daughters by their respective volatilities give strongclues on the timing of terrestrial differentiation and the early cycle of volatileelements. Noble gases are chemically inert and their isotopic composition canonly be modified by kinetic fractionation, or mixing with nucleosyntheticcomponents, or through nuclear reactions including extinct radioactivitydecays.

Xenon, the heaviest stable noble gas, is of particular interest because someof its isotopes are the radioactivity products of three different decay systemscovering contrasted time intervals. Iodine is a volatile element for whichone isotope, 129I decays with T1/2 = 15.7 Myr to 129Xe. During terrestrialmagma ocean episodes, xenon, which has presumably a higher volatility thaniodine, was degassed preferentially to it. The amount of 129Xe in the atmo-sphere, in excess of the non-radiogenic xenon composition, corresponds to a‘‘closure’’ interval of about 100 Myr (Allegre et al., 1995). Put in otherwords, only a tiny fraction of radiogenic 129Xe has been retained in theatmosphere, showing evidence that the atmosphere was open to loss in spacefor at least several tens of Myr. The terrestrial mantle has kept even less129Xe. Heavy xenon isotopes e.g., 136Xe, are produced by the spontaneousfission of 244Pu (T1/2 = 82 Myr) so that the combination of both chro-nometers allows one to date the closure of the terrestrial mantle at60–70 Myr (Kunz et al., 1998). This age could be interpreted as averagingthe period during which the magma ocean episodes declined enough toquantitatively retain is volatile elements. Notably, heavy Xe isotopes are alsoproduced by the spontaneous fission of 238U which decays with a half-life of4.45 Gyr, so that it is possible to compute closure ages based on244Pu–238U–136Xe. Results indicate a closure age of about 400–600 Myr


(Yokochi and Marty, 2005), which could correspond to a significant declinein mantle geodynamics. Xe isotope variations are therefore consistent withlarge-scale differentiation within the first 100 Myr and also with prolongedmantle activity at rates much higher than at Present during the Hadean. Acomparable scenario was derived from other radioactive tracers, notably the146Sm–142Nd and 147Sm–143Nd systems, and this view is fully consistent withmodels linking the thermal state of the Earth with the rate of mantleconvection.

3.3.3.2. Geological evidence: Core differentiation, magnetic fieldThe time it took to build the Earth and the other terrestrial planets has beeninvestigated from two different approaches, theoretical modeling on onehand (see previous sections), and absolute chronology on the other hand. Inshort, numerical simulations indicate that the growth of terrestrial planetswas a geologically fast process. In the turbulent nascent solar system, dustaccreted into small bodies, for which gravity became the main coalescentagent, within 104–105 years. Models predict that bodies with sizes up to thatof Mars accreted within a few Myr. Independent evidence for rapid growthstems from coupled variations of 142Nd, 182W, 129Xe and 131–136Xe producedby extinct radioactivities of 144Sm (T1/2 = 106 Myr), 182Hf (9 Myr), 129I(15.7 Myr) and 244Pu (82 Myr) observed in Martian meteorites (Hallidayet al., 2001; Marty and Marti, 2002). These geochemical tracers attest thatMartian differentiation including core formation, crustal development andmantle degassing took place within £15 Myr, which is remarkably short, andcontemporaneous to the differentiation of parent bodies of meteorites.Furthermore, they also indicate that these heterogeneities were notre-homogenized later on as is the case of the Earth, and therefore attest for alow mantle convection rate on Mars, if any.

As dicussed previously, it took about 100 Myr to complete the Earth,considerably longer than the time of Mars-sized object formation (Wether-hill, 1980). This longer time interval is the result of the decreasing probabilityfor collisions to occur in an increasingly depleted population of growingbodies. This modeling approach does not allow one to get more detailedchronological definition within this time frame, which needs to be con-strained independently. The tungsten isotopic composition of the Earthcompared to that of the nascent solar system as recorded in primitivemeteorites supports large-scale differentiation within a few tens of millionyears. 182Hf decays into 182W with a half-life T1/2 = 9 Myr. Duringmetal–silicate differentiation, tungsten is siderophile, that is, concentratespreferentially in metal phase whereas hafnium prefers the silicates (litho-phile). If, in a differentiated body, this event took place when 182Hf was still


present, the isotopic composition of tungsten is different from the solar oneas recorded for example in undifferentiated meteorites.

In pioneering attempts to use this chronometer, no difference in thetungsten isotopic composition was found between carbonaceous chondrites(the most primitive meteorites found so far) and terrestrial silicates, leadingthe authors to conclude that the last global Hf–W differentiation of the Earthhappened after all 182W decayed, in practice ‡60 Myr after the start of solarsystem condensation t0 (Lee and Halliday, 1995). More recently, differencesbetween chondrites and the Earth have been found (Kleine et al., 2002; Yinet al., 2002), implying a mean metal–silicate differentiation of 30 Myr for theEarth if terrestrial differentiation was a single and global event, with a pos-sible range of 11–50 Myr for more realistic accreting conditions. A collisionbetween a Mars-sized object and a growing proto-Earth is consistent with theunique Earth–Moon angular momentum. Numerical models for such a giantimpact indicate that the proto-Earth was severely disrupted while materialfrom the proto-mantle and the impactor spiraled at high temperature andformed the Moon (see above, Section 3.3.2). It is therefore logical, even if ithas not yet been demonstrated, to ascribe to this collision the last globaldifferentiation between metal and silicate recorded in the Hf–W system.Recent tungsten data for lunar basalts indicate a 182W anomaly for the lunarmantle, interpreted as record for differentiation at 45±4 Myr (Kleine et al.,2005). These authors proposed that this age represented the end of magmaocean episodes on the Moon.

The earliest record of a geomagnetic field dates back to the Archean (Haleand Dunlop, 1984; Yoshihara and Hamano, 2004), some 1 Gyr after theformation period. The terrestrial magnetic field has a major role in pre-venting solar wind ions to interact extensively with the top of the atmosphere,and create isotope fractionation of atmospheric elements by pick-up chargeexchange. It also tends to preserve the surface of the Earth from cosmic-raybombardment that are lethal for the development of organic chemistry.However, there is no evidence for the existence of magnetic field induced bythe geodynamo during the first Gyr.

References

Abergel, A. et al.: 1996, Astron. Astrophys. 315, L329–L332.Adams, F. C. and Myers, P. C.: 2001, Astrophys. J. 553, 744–753.Adams, F. C., Hollenbach, D., Laughlin, G. and Gorti, U.: 2004, Astrophys. J. 611, 360–379.

Adams, F. C., Proszkow, E. M., Fatuzzo, M., and Myers, P. C.: 2006, Astrophys. J. 641,504–525.

Agnor, C. B., Canup, R. M. and Levison, H. F.: 1999, Icarus 142, 219–237.

Alibert, Y., Mordasini, C., Benz, W. and Winisdoerffer, C.: 2005, Astorn. Astrophys. 434,343–353.


Allegre, C. J., Manhes, G. and Gopel, C.: 1995, Geochim. Cosmochim. Acta. 59, 1445–1456.Amelin, Y., Krot, A. N., Hutcheon, I. D. and Ulyanov, A. A.: 2002, Science 297, 1678–1683.

Anders, E. and Grevesse, N.: 1989, Geochim. Cosmochim. Acta 53, 197–214.Andre, P. and Montmerle, T.: 1994, Astrophys. J. 420, 837–862.Andre, P., Ward-Thompson, D. and Barsony, M.: 2000, in V. Mannings, A. P. Boss and S. S.

Russell (eds.), Protostars and Planets IV, University of Arizona Press, Tucson, pp. 59–97.Baker, J., Bizzarro, M., Wittig, N., Connelly, J. and Haack, H.: 2005, Nature 436, 1127–1131.Balbus, S. A. and Hawley, J. F.: 1991, Astrophys. J. 376, 214–233.Bate, M. R. and Bonnell, I. A.: 2004, in H. J. G. L. M. Lamers, L. J. Smith, and A. Nota (eds.),

The Formation and Evolution of Massive Young Star Clusters. ASP Conf. Ser., 322, 289.Beaulieu, J.-Ph. et al.: 2006, Nature 439, 437–440.Benz, W.: 2000, Sp. Sci. Rev. 92, 279–294.

Bertout, C.: 1989, Ann. Rev. Astron. Astrophys. 27, 351–395.Bizzarro, M., Baker, J. A. and Haack, H.: 2004, Nature 431, 275–278.Boss, A. P.: 2003, Astrophys. J. 599, 577–581.

Bouvier, A., Blichert-Toft, J., Vervoort, J. D., McClelland, W. and Albarede, F.: 2005,Geochim. Cosmochim. Acta Suppl. 69(Suppl. 1), A384.

Bouvier, J. et al.: 1999, Astron. Astrophys. 349, 619–635.

Bouvier, J. et al.: 2003, Astron. Astrophys. 409, 169–192.Brearley, A. and Jones, R. H.: 1998, in J. J. Papike (ed.), Planetary materials. Rev. Mineral.

36, 3/1–3/398.Busso, M., Gallino, R. and Wasserburg, G. J.: 2003, Pub. Astr. Soc. Austr. 20, 356–370.

Canup, R. M.: 2004a, Icarus 168, 433–456.Canup, R. M.: 2004b, Ann. Rev. Astron. Astrophys. 42, 441–475.Canup, R. M. and Asphaug, E.: 2001, Nature 412, 708–712.

Canup, R. M., Levison, H. F. and Stewart, G. R.: 1999, Astr. J. 117, 603–620.Canup, R. M. and Esposito, L. W.: 1996, Icarus 119, 427–446.Carpenter, J. M., Wolf, S., Schreyer, K., Launhardt, R. and Henning, T.: 2005, Astr. J. 129,

1049–1062.Chambers, J. E. and Wetherill, G. W.: 1998, Icarus 136, 304–327.Chambers, J. E. and Wetherill, G. W.: 2001, Met. Plan. Sci. 36, 381–399.

Chambers, J. E.: 2001, Icarus 152, 205–224.Chaussidon, M. and Gounelle, M.: 2006, in D. Lauretta and L. Leshin (eds.), Meteorites &

Early Solar System II. Tucson, University of Arizona Press, pp. 323–340.Chaussidon, M., Libourel, G. and Krot, A. N.: 2006, Lunar Planet. Sci. 36, 1335.

Chaussidon, M., Robert, F. and McKeegan, K. D.: 2006, Geochim. Cosmochim. Acta 70, 224–245.

Chokshi, A., Tielens, A. G. G. M. and Hollenbach, D.: 1993, Astrophys. J. 407, 806–819.

Clark, P. C., Bonnell, I. A., Zinnecker, H. and Bate, M. R.: 2005, Mon. Not. Roy. Astr. Soc.359, 809–818.

Crutcher R. M.: 2005, in Magnetic fields in the Universe: From Laboratory and Stars to

Primordial Structures, AIP Conf. Proc., vol. 784, pp. 129–139.Diehl, R., Halloin, H., Kretschmer, K., Lichti, G. G., Schonfelder, V., Strong, A. W., von

Kienlin, A., Wang, W., Jean, P., Knodlseder, J., et al.: 2006, Nature 439, 45–47.Diehl, R., Cervino, M., Hartmann, D. H. and Kretschmar, K.: 2004, New Astr. Rev. 48, 81–86.

Donati, J.-F., Paletou, F., Bouvier, J. and Ferreira, J.: 2005, Nature 438, 466–469.Draine, B. T.: 2006, Astrophys. J. 636, 1114–1120.Duchene, G., McCabe, C., Ghez, A. M. and Macintosh, B. A.: 2004, Astrophys. J. 606, 969–

982.Ehrenfreund, P. and Charnley, S. B.: 2000, Ann. Rev. Astr. Ap. 38, 427–483.


Falgarone, E., Hily-Blant, P., Pety, J. and Pineau Des Forets, G.: 2005 in Magnetic fields inthe Universe: From Laboratory and Stars to Primordial Structures, AIP Conf. Proc., vol.

784, pp. 299–307.Favata, F., Micela, G., Silva, B., Sciortino, S. and Tsujimoto, M.: 2005, Astron. Astrophys.

433, 1047–1054.

Feigelson, E. D.: 2005, Met. Planet. Sci. 40(Suppl.), 5339.Feigelson, E. D. and Montmerle, T.: 1999, Ann. Rev. Astron. Astrophys. 37, 363–408.Feigelson, E. D., Getman, K., Townsley, L., Garmire, G., Preibisch, T., Grosso, N., Mont-

merle, T., Muench, A. and McCaughrean, M.: 2005, Astrophys. J. Suppl. Ser. 160, 379–

389.Ferreira, J., Pelletier, G. and Appl, S.: 2000, Mon. Not. R. Astron. Soc. 312, 387–397.Ferreira, J. and Casse, F.: 2004, Astrophys. J. 601, L139–L142.

Fleming, T. and Stone, J. M.: 2003, Astrophys. J. 585, 908–920.Fromang, S., Balbus, S. A., Terquem, C. and De Villiers, J.-P.: 2004, Astrophys. J. 616, 364–

375.

Fromang, S., Terquem, C. and Balbus, S. A.: 2002, Mon. Not. R. Astron. Soc. 329, 18–28.Gallino, R., Busso, M., Wasserburg, G. J. and Straniero, O.: 2004, New Astr. Rev. 48, 133–

138.

Galy, A., Young, E. D., Ash, R. D. and O’Nions, R. K.: 2002, Science 290, 1751–1753.Getman, K. V., Feigelson, E. D., Grosso, N., McCaughrean, M. J., Micela, G., Broos, P.,

Garmire, G. and Townsley, L.: 2005, Astrophys. J. Suppl. Ser. 160, 353–378.Glassgold, A. E., Feigelson, E. D. and Montmerle, T.: 2000, in V. Mannings, A. P. Boss and

S. S. Russell (eds.), Protostars and Planets IV, University of Arizona Press, Tucson, pp.429–455.

Glassgold, A. E., Feigelson, E. D., Montmerle, T. and Wolk, S.: 2005, in A. N. Krot, E. R. D.

Scott and B. Repurth (eds.), Chondrites and the Protoplanetary Disk. ASP Conf. Ser., vol.341, pp. 165–182.

Goldreich, P. and Ward, W. R.: 1973, Astrophys. J. 183, 1051–1061.

Goodwin, S. P., Whitworth, A. P. and Ward-Thompson, D.: 2004, Astron. Astrophys. 423,169–182.

Gounelle, M. and Russell, S. S.: 2005, Geochim. Cosmochim. Acta 69, 3129–3144.

Gounelle, M., Shu, F. H., Shang, H., Glassgold, A. E., Rehm, K. E. and Lee, T.: 2004, LunarPlanet. Sci. Conf. 35, 1829.

Gounelle, M., Shu, F. H., Shang, H., Glassgold, A. E., Rehm, K. E. and Lee, T.: 2001,Astrophys. J. 548, 1051–1070.

Greaves, J. S.: 2005, Mon. Not. R. Astron. Soc. Letters 364, L47–L50.Greenberg, R., Wacker, J. F., Hartmann, W. K. and Chapman, C. R.: 1978, Icarus 35, 1–26.Grosso, N., Alves, J., Wood, K., Neuhauser, R., Montmerle, T. and Bjorkman, J. E.: 2003,

Astrophys. J. 586, 296–305.Guillot, T.: 1999, Science 286, 72–77.Guillot, T.: 2005, Ann. Rev. Earth Plan. Sci. 33, 493–530.

Guillot, T. and Hueso, R.: 2006, Mon. Not. R. Astron. Soc. 367, L47–L51.Guilloteau, S., Dutrey, A. and Simon, M.: 1999, Astr. Astrophys. 348, 570–578.Haisch, K. E., Lada, E. A. and Lada, C. J.: 2001, Astrophys. J. Lett. 553, L153–L156.Hale, C. J. and Dunlop, D.: 1984, Geophys. Res. Lett. 11, 97–100.

Halliday, A. N., Wanke, H., Birck, J. L. and Clayton, R. N.: 2001, Space Sci. Rev. 96, 1–34.Harker, D. E. and Desch, S. J.: 2002, Astrophys. J. Lett. 565, L109–L112.Harris, M. J., Knodlseder, J., Jean, P., Cisana, E., Diehl, R., Lichti, G. G., Roques, J.-P.,

Schanne, S. and Weidenspointner, G.: 2005, Astron. Astrophys. 433, L49–L52.Hayashi, C.: 1966, Ann. Rev. Astron. Astrophys. 4, 171–192.


Hayashi, C.: 1981, Prog. Theor. Phys. 70(Suppl.), 35–53.Hayashi, M., Shibata, K. and Matsumoto, R.: 2000, Adv. Sp. Res. 26, 567–570.

Heiles, C., Haffner, L. M., Reynolds, R. J. and Tufte, S. L.: 2000, Astrophys. J. 536, 335–346.Hester, J. J. and Desch, S.: 2005, in A. N. Krot and E. R. D. Scott and B. Reipurth (eds.),

Chondrites and the Protoplanetary Disk. ASP Conf series, vol. 341, pp. 107–130.

Hewins, R. H.: 1997, Ann. Rev. Earth Planet. Sci. 25, 61–83.Hewins, R. H., Connolly, H. C. Jr., Lofgren, G. E. and Libourel, G.: 2005, in A. N. Krot, E.

R. D. Scot and B. Reipurth (eds.), Chondrites and the Protoplanetary Disk. ASP Conf.Series, vol. 341, pp. 286–316.

Hillenbrand, L. A.: 1997, Astron. J. 113, 1733–1768.Hillenbrand, L. A.: 2006, in Livio M. (ed.), A Decade of Discovery: Planets Around Other

Stars. STScI Symposium Series 19, in press.

Huss, G. R. and Tachibana, S.: 2004, Lunar Planet. Sci. Conf. 35, 1811.Ida, S., Canup, R. M. and Stewart, G. R.: 1997, Nature 389, 353–357.Ida, S. and Makino, J.: 1993, Icarus 106, 210–227.

Ilgner, M. and Nelson, R. P.: 2006, Astron. Astrophys. 445, 205–222.Johns-Krull, C. M. and Gafford, A. D.: 2002, Astrophys. J. 573, 685–698.Kessler-Silacci, J. E., Augereau, J.-C. and Dullemond, C. P. et al.: 2006, Astrophys. J. 639,

275–291.Kleine, T., Metzger, K. and Palme, H.: 2005, Lunar Planet. Sci. 36, 1940, CD-ROM.Kleine, T., Munker, C., Mezfer, K. and Palme, H.: 2002, Nature 418, 952–955.Kokubo, E. and Ida, S.: 1998, Icarus 131, 171–178.

Kokubo, E., Ida, S. and Makino, J.: 2000, Icarus 148, 419–436.Kroupa, P.: 2002, Science 295, 82–91.Krumholz, M. R., McKee, C. F. and Klein, R. I.: 2005, Nature 438, 332–334.

Kunz, J., Staudacher, T. and Allegre, C. J.: 1998, Science 280(5365), 877–880.Laskar, J. and Robutel, P.: 1993, Nature 361, 608–612.Lee, D. C. and Halliday, A.: 1995, Nature 378, 771–774.

Lee, T., Shu, F. H., Shang, H., Glassgold, A. E. and Rhem, K. E.: 1998, Astrophys. J. 506,898–912.

Lissauer, J.: 1987, Icarus 69, 249–265.

Lyo, A.-Ran, Lawson, W. A., Mamajek, E. E., Feigelson, E. D., Sung, Eon-Chang andCrause, L. A.: 2003, Mon. Not. R. Astron. Soc. 338, 616–622.

MacPherson, G. J., Davis, A. M. and Zinner, E. K.: 1995, Meteoritics 30, 365–386.MacPherson, G. J. and Huss, G. R.: 2003, Geochim. Cosmochim. Acta 67, 3165–3179.

Matsumura, S. and Pudritz, R. E.: 2006, Mon. Not. R. Astron. Soc. 365, 572–584.Matsumoto, R., Machida, M., Hayashi, M. and Shibata, K.: 2000, Progr. Th. Phys.

138(Suppl.), 632–637.

Matzner, C. D. and McKee, C. F.: 2000, Astrophys. J. 545, 364–378.Megeath, S. T., Hartmann, L., Luhman, K. L. and Fazio, G. G.: 2005, Astrophys. J. Lett. 634,

L113–L116.

Manhes G., Gopel C. and Allegre C. J.: 1988, Comptes Rendus de l’ATP Planetologie, 323–327.

Marty, B. and Marti, K.: 2002, Earth Plan. Sci. Letts. 196(3–4), 251–263.Masset, F., Morbidelli, A., Crida, A. and Ferreira, J.: 2006, Astrophys. J. 642, 478–487.

Masset, F. and Snellgrove, M.: 2001, Mon. Not. R. Astron. Soc. 320, L55–L59.McKeegan, K. D. and Davis, A. M.: 2003, in H. Holl and K. Turekian (eds.), Treatise on

Geochemistry, Elsevier-Pergamon, Oxford, pp. 431–460.

McKeegan, K. D., Chaussidon, M. and Robert, F.: 2000, Science 289, 1334–1337.


Meyer, B. S.: 2005, in A. N. Krot, E. R. D. Scott and B. Reipurth (eds.), Chondrites and theProtoplanetary Disk, ASP Conf series, vol. 341, pp. 515–526.

Micela, G. and Favata, F.: 2005, Sp. Sci. Rev. 108, 577–708.Montmerle, T.: 1979, Astrophys. J. 231, 95–110.Montmerle, T.: 2002, New Astr. Rev. 46, 573–583.

Montmerle, T.: 2005, in M. Gargaud, B. Barbier, H. Martin and J. Reisse (eds.), Lectures inAstrobiology, Springer, Heidelberg, pp. 27–59.

Morbidelli, A. and Levison, H. F.: 2004, Astron. J. 128, 2564–2576.Mostefaoui, S., Kita, N. T., Togashi, S., Tachibana, S., Nagahara, H. and Morishita, Y.:

2002, Meteoritics Planet. Sci. 37, 421–438.Mostefaoui, S., Lugmair, G. W., Hoppe, P. and El Goresy, A.: 2004, New Astron. Rev. 48,

155–159.

Motte, F., Andre, P. and Neri, R.: 1998, Astron. Astrophys. 336, 150–172.Muzerolle, J., Calvet, N. and Hartmann, L.: 2001, Astrophys. J. 550, 944–961.Natta, A., Testi, L., Calvet, N., Henning, Th., Waters, R. and Wilner, D.: 2006, in: B.

Reipurth (ed.), Protostars and Planets V. Tucson, University of Arizona Press, in press(eprint arXiv:astro-ph/0602041).

Nelson, R. P.: 2005, Astron. Astrophys. 443, 1067–1085.

Ozawa, H., Grosso, N. and Montmerle, T.: 2005, Astron. Astrophys. 429, 963–975.Padoan, P., Jimenez, R., Juvela, M. and Nordlund, A.: 2004, Astrophys. J. 604, L49–L52.Pahlevan, K. and Stevenson, D. J.: 2005, Lunar Planet. Sci. Conf. 36, 1505.Palla, F. and Stahler, S. W.: 1999, Astrophys. J. 525, 772–783.

Petit, J.-M., Morbidelli, A. and Chambers, J.: 2001, Icarus 153, 338–347.Pety, J. and Falgarone, E.: 2003, Astron. Astrophys. 412, 417–430.Podolak, M.: 2003, Icarus 165, 428–437.

Podosek, F. A., Zinner, E. K., MacPherson, G. J., Lundberg, L. L., Brannon, J. C. and Fahey,A. J.: 1991, Geochim. Cosmochim. Acta 55, 1083–1110.

Pollack, J. B., Hubickyj, O., Bodenheimer, P., Lissauer, J. J, Podolak, M. and Greenzweig, Y.:

1996, Icarus 124, 62–85.Poppe, T. and Blum, J.: 1997, Adv. Space. Res. 20, 1595–1604.Poppe, T., Blum, J. and Henning, T.: 2000, Astophys. J. 533, 454–471.

Reipurth, B. and Bally, J.: 2001, Ann. Rev. Astron. Astrophys. 39, 403–455.Ryter, C.: 1996, Astr. Sp. Sci. 236, 285–291.Sasaki, T. and Abe, Y.: 2004, Lunar Planet. Sci. Conf. 35, 1505.Shang, H., Glassgold, A. E., Shu, F. H. and Lizano, S.: 2002, Astrophys. J. 564, 853–876.

Shu, F. H., Adams, F. C. and Lizano, S.: 1987, Ann. Rev. Astron. Astrophys. 25, 23–81.Shu, F. H., Shang, H., Glassgold, A. E. and Lee, T.: 1997, Science 277, 1475–1479.Stelzer, B. and Neuhauser, R.: 2001, Astron. Astrophys. 377, 538–556.

Stolper, E. and Paque, J. M.: 1986, Geochim. Cosmochim. Acta 50, 1785–1806.Tanga, P., Babiano, A., Dubrulle, B. and Provenzale, A.: 1996, Icarus 121, 158–177.Thommes, E. W., Duncan, M. J. and Levison, H. F.: 2003, Icarus 161, 431–455.

Throop, H. B. and Bally, J.: 2005, Astrophys. J. 623, L149–L152.Tsujimoto, M., Feigelson, E. D., Grosso, N., Micela, G., Tsuboi, Y., Favata, F., Shang, H.

and Kastner, J. H.: 2005, Astrophys. J. Suppl. Ser. 160, 503–510.Wasserburg, G. J.: 1987, Earth Plan. Sci. Lett. 86, 129–173.

Weidenschilling, S. J.: 1977, Mon. Not. R. Astron. Soc. 180, 57–70.Weidenschilling, S. J.: 1980, Icarus 44, 172–189.Wetherhill, G. W.: 1980, Ann. Rev. Astron. Astrophys. 18, 77.

Wetherhill, G. W.: 1990, Ann. Rev. of Earth Plan. Sciences 18, 205–256.


Yin, Q., Jacobsen, S. B., Yamashita, K., Telouk, P., Blichert-Toft, J. and Albarede, F.: 2002,Nature 418, 949–952.

Yokochi, R. and Marty, B.: 2005, Earth Plan. Sci. Letts 238, 17–30.Young, E. D., Simon, J. I., Galy, A., Russell, S. S., Tonui, E. and Lovera, O.: 2005, Science

308, 223–227.

Yoshihara, A. and Hamano, Y.: 2004, Precambr. Res. 131, 111–142.Zanda, B.: 2004, Earth Planet. Sci. Lett. 224, 1–17.


Documents

THIERRY MONTMERLE and JEAN-CHARLES AUGEREAU · (like the Sun today) vs. the number of stars in star-forming regions leads to a probability argument drawn from observations: nearly